feat: add daily paper cookbook and DingTalk agent integration (#385)

* feat(daily-paper): add daily paper cookbook workflow with schema and tests - Introduce daily paper schema types (DailyBriefOutput, PaperInfo, PaperNoteOutput, etc.) - Create daily paper cookbook module with analyze, collect, digest, rank, and select steps - Add cookbook entry point and integrate into main steps module - Replace job config export with daily brief output in schema exports - Add comprehensive unit tests covering pipeline, filtering, and output generation - Update dependencies including openai-codex and pypdf packages - Configure standalone daily paper cron job with proper scheduling and routing * test(daily_paper): update tests to use Claude Code wrapper exclusively - Add test to verify web search is disallowed by default in Claude Code - Update imports to include DailyBriefOutput, PaperNoteOutput, and PaperSelection schemas - Change test name from standalone_config_has_backend_split to reflect Claude Code only usage - Remove default agent wrapper and configure all steps to use Claude Code wrapper - Rename select_wrapper to cc_wrapper for clarity and consistency - Remove duplicate Claude Code wrapper initialization - Update test assertions to verify output schema usage matches expected sequence - Remove unused as_llm component from standalone configuration test * refactor(agent-wrapper): simplify skill resolution logic across all wrappers - Replace duplicate skill resolution code with centralized _resolve_project_skills method - Add project_path property with configurable relative path resolution - Introduce proper validation for skill names and directory existence - Change Codex wrapper to use project_path instead of workspace_path for skills - Add SKILL.md requirement validation for project skills - Remove redundant skill processing logic from individual wrappers * feat(daily_paper): add daily paper workflow with PDF analysis and brief generation - Implement shared state management and file helpers for daily-paper steps - Add PDF download and text extraction capabilities with arXiv integration - Create paper collection step with Hugging Face weekly/monthly rankings - Build ranking system using reciprocal-rank fusion with memory keyword scoring - Add Claude Code integration for paper analysis and detailed note generation - Implement digest step to create final five-minute brief from detailed notes - Add configuration for standalone daily cookbook application with cron scheduling - Create typed schema for paper information, selection, and output formats - Add atomic file writing with temporary file safety mechanisms - Implement exclusion logic for previously recommended papers and daily filters * feat(daily_paper): add DingTalk notification integration and enhance logging - Integrate DingTalk markdown send step to notify groups about daily paper briefs - Add comprehensive logging throughout daily paper workflow including start/finish events - Update daily paper analysis prompt to include code repository context requirement - Configure DingTalk notification in daily_cookbook.yaml with app credentials - Add dingtalk-stream dependency for proactive message API integration - Enhance daily paper README with DingTalk notification section and updated flow chart - Implement detailed logging for each step including paper processing and agent calls - Add test coverage for DingTalk markdown sending functionality and configuration - Update pre-commit config to exclude skills directory from checks - Add .claude/skills to gitignore for local development environment * refactor(dingtalk): move dingtalk_stream import to local scope and improve code safety - Moved global dingtalk_stream import to local scope in send.py to avoid eager loading - Added dynamic import with error handling for optional dependency cases - Updated test suite to verify lazy loading behavior works correctly - Fixed markdown title generation by using safe variable naming in wait.py - Enhanced test coverage for arxiv PDF download caching functionality - Updated application context initialization with proper resource directory configuration - Modified paper metadata to include source PDF path reference in output files * refactor(daily_paper): remove manifest system and store selection metadata in digest files - Remove JSON manifest creation and storage functionality - Store selection data directly in digest file frontmatter instead of separate manifest files - Add load_saved_selection method to rebuild selection from digest and paper-note metadata - Update README documentation to reflect new cookbook workflow architecture - Modify test cases to verify selection metadata in digest files instead of manifest JSON - Remove unused json import from multiple daily paper modules - Integrate PaperSelection schema for proper data validation in stored metadata * docs(daily_paper): add bilingual cookbook guides
2026-08-28 05:25:04 +00:00 · 2026-07-22 19:17:01 +08:00 · 2026-07-22 19:17:01 +08:00 · 46adb5ae1e
commit 46adb5ae1e
parent 630f26b119
40 changed files with 6399 additions and 60 deletions
--- a/.gitignore
+++ b/.gitignore
@ -54,6 +54,12 @@ vault/
 docs/_build/
 site/

+evaluation/
+datasets/
+
+# Claude Code skills (local only)
+.claude/skills/
+
 # Memory workspaces (keep dirs, ignore contents)
 benchmark/memory_workspaces/

--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@ -1,3 +1,5 @@
+exclude: ^skills/
+
 repos:
  - repo: https://github.com/pre-commit/pre-commit-hooks
    rev: v6.0.0
--- a/README.md
+++ b/README.md
@ -58,6 +58,9 @@ memory, then continuously indexes, links, and consolidates that memory for futur

 ## 📰 News

+- [2026.07] - Introduced optional Cookbook workflows, starting with
+  [Daily Paper](cookbook/daily_paper/README.md) for scheduled paper discovery, agent-assisted PDF analysis, reusable
+  Markdown notes, and five-minute briefs.
 - [2026.07] - Our
  paper [Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution](https://aclanthology.org/2026.findings-acl.829/)
  has been accepted to Findings of ACL 2026.
@ -161,6 +164,16 @@ ReMe stores agent memory as readable Markdown.
 Related: [[digest/wiki/memory-as-file.md]]
 ```

+## 🧑‍🍳 Cookbooks
+
+Cookbooks are optional, end-to-end workflows assembled from ReMe jobs and steps. They are not enabled by the default
+configuration; select the cookbook's standalone configuration when starting ReMe. Each new cookbook will be added as
+another row in this table.
+
+| Cookbook    | Capability                                                                                                    | Introduction                             |
+|-------------|---------------------------------------------------------------------------------------------------------------|------------------------------------------|
+| Daily Paper | Discover and rank papers, analyze PDFs with an agent, and generate file-native notes and a five-minute brief. | [README](cookbook/daily_paper/README.md) |
+
 ## 📁 Memory System

 > Memory as File, File as Memory.
--- a/README_ZH.md
+++ b/README_ZH.md
@ -51,6 +51,8 @@ Agent 能够可靠召回。

 ## 📰 新闻

+- [2026.07] - 新增可选的 Cookbook 工作流，首个能力为 [每日论文](cookbook/daily_paper/README_ZH.md)，支持定时发现论文、
+  Agent 辅助解析 PDF、沉淀可复用的 Markdown 笔记并生成五分钟简报。
 - [2026.07] -
  我们的论文 [Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution](https://aclanthology.org/2026.findings-acl.829/)
  已被 Findings of ACL 2026 接收。
@ -154,6 +156,15 @@ ReMe 会把 Agent 记忆保存为可读的 Markdown。
 相关链接：[[digest/wiki/memory-as-file.md]]
 ```

+## 🧑‍🍳 Cookbooks
+
+Cookbook 是由 ReMe jobs 和 steps 组装而成的可选端到端工作流。默认配置不会开启它们；启动 ReMe 时选择对应的
+独立配置即可启用。后续新增的 cookbook 会继续在表格中按行追加。
+
+| Cookbook | 能力                                       | 介绍                                        |
+|----------|------------------------------------------|-------------------------------------------|
+| 每日论文     | 发现并排序论文，使用 Agent 解读 PDF，生成文件化论文笔记和五分钟简报。 | [使用说明](cookbook/daily_paper/README_ZH.md) |
+
 ## 📁 记忆系统

 > Memory as File, File as Memory.
@ -277,7 +288,7 @@ frontmatter 和文件操作接口主要用于维护、调试或高级集成。
 |-------------------------------------------|---------------------------------------------|
 | `reme start`                              | 启动本地 ReMe 服务。                               |
 | `reme version` / `reme health_check`      | 检查包版本和组件状态。                                 |
-| `reme status`                             | 查看有状态数据组件的内存估算及进程 RSS。               |
+| `reme status`                             | 查看有状态数据组件的内存估算及进程 RSS。                      |
 | [`reme search`](docs/zh/memory_search.md) | 默认使用 BM25 和 wikilink 检索，启用后增加向量检索。          |
 | `reme read` / `reme write` / `reme edit`  | 检查和维护 Markdown 记忆文件。                        |
 | `reme auto_memory`                        | 将对话 messages 转为 daily 记忆卡片；需要 LLM 凭证。       |
--- a/cookbook/daily_paper/README.md
+++ b/cookbook/daily_paper/README.md
@ -0,0 +1,297 @@
+# Daily Paper Cookbook
+
+[中文](README_ZH.md)
+
+Daily Paper is a local-first, file-native workflow for turning research feeds into a daily reading package. It collects
+papers from the Hugging Face weekly and monthly rankings, removes yesterday's papers and recent recommendations, ranks
+the remaining candidates, and uses Claude Code to produce detailed Chinese paper notes and a five-minute Chinese brief.
+
+The workflow is assembled by [`daily_cookbook.yaml`](../../reme/config/daily_cookbook.yaml). Its schemas live in
+[`reme/schema/daily_paper.py`](../../reme/schema/daily_paper.py), and its steps live in
+[`reme/steps/cookbook/daily_paper/`](../../reme/steps/cookbook/daily_paper/).
+
+## Quick start
+
+Daily Paper requires Python 3.11 or later, the `core` dependencies, network access to Hugging Face and arXiv, and
+credentials for the configured Claude Code endpoint.
+
+From the repository root:
+
+```bash
+python -m pip install -e ".[dev,core]"
+export CLAUDE_CODE_API_KEY="your-api-key"
+reme start config=daily_cookbook job=daily_paper
+```
+
+The built-in configuration uses `qwen3.7-max` through DashScope's Anthropic-compatible endpoint. Override
+`CLAUDE_CODE_MODEL_NAME` and `CLAUDE_CODE_BASE_URL` when using another compatible model or provider.
+
+By default, outputs are written under `.reme/` in the directory where ReMe starts.
+
+## What it creates
+
+A successful run writes ordinary PDFs and Markdown files beneath `workspace_dir`:
+
+```text
+.reme/
+├── daily/
+│   ├── YYYY-MM-DD.md
+│   └── YYYY-MM-DD/
+│       ├── daily-paper-brief.md
+│       ├── paper-<arxiv-id>.md
+│       └── ...
+├── resource/
+│   └── papers/
+│       ├── <arxiv-id>.pdf
+│       └── ...
+└── mem_session/
+    └── claude_config/
+```
+
+- `paper-<arxiv-id>.md` is a detailed Chinese reading note with YAML frontmatter linking back to the source PDF and
+  paper pages.
+- `daily-paper-brief.md` is a roughly five-minute Chinese digest with wikilinks to every selected paper note.
+- `daily/YYYY-MM-DD.md` is a derived day index rebuilt from the Markdown files for that date.
+- `resource/papers/` holds reusable source PDFs.
+
+The paper notes are the source of truth for recommendation history: their frontmatter contains the `arxiv_id` values
+used for future deduplication. The day index is derived and can be rebuilt. The workflow does not currently write a
+separate run manifest.
+
+## How the workflow works
+
+```mermaid
+flowchart LR
+    HF[Hugging Face<br/>weekly + monthly] --> C[1. Collect]
+    Y[Yesterday's papers] --> C
+    H[Recent local notes] --> C
+    C --> R[2. Rank]
+    R --> S[3. Select]
+    S --> A[4. Analyze PDFs]
+    A --> D[5. Build digest]
+    D --> N[6. Notify DingTalk]
+    A --> P[PDFs + paper notes]
+    D --> B[Brief + day index]
+```
+
+### 1. Collect and deduplicate
+
+The Collect step fetches the weekly ranking for the run date's ISO week, the monthly ranking for its calendar month,
+and the Hugging Face Daily Papers IDs for exactly the previous calendar day. It merges weekly and monthly metadata by
+arXiv ID and preserves each list's display rank.
+
+It then scans `daily/<prior-date>/paper-*.md` over the configured history window and excludes IDs found in note
+frontmatter. The job fails clearly if no eligible papers remain.
+
+### 2. Rank candidates
+
+The Rank step uses reciprocal-rank fusion:
+
+```text
+score = 1 / (rrf_k + monthly_rank)
+      + weekly_weight / (rrf_k + weekly_rank)
+```
+
+A missing rank contributes zero. Candidates are ordered by fused score, upvotes, and arXiv ID. The bounded candidate
+pool also reserves several positions for papers whose titles or summaries match memory-related terms such as agent
+memory, memory retrieval, continual learning, context compression, knowledge graphs, and RAG. This reserve is a simple
+keyword heuristic, not a semantic classifier.
+
+### 3. Select papers
+
+Claude Code receives the bounded candidate pool and returns a structured `PaperSelection`. The implementation requires
+exactly `top_k` unique in-pool IDs with consecutive ranks. Invalid output is returned to the agent once as validation
+feedback; a second invalid response fails the job.
+
+### 4. Download and analyze PDFs
+
+Selected papers are processed sequentially. For each paper, the workflow:
+
+1. validates the modern arXiv ID format;
+2. downloads and validates the PDF, or reuses an existing file with a valid `%PDF-` header;
+3. extracts text with `pypdf`, adding page markers and applying page and character limits;
+4. asks Claude Code for a structured detailed reading; and
+5. writes normalized frontmatter plus the generated Markdown body.
+
+The current extractor requires a usable PDF text layer. Scanned or image-only PDFs fail because there is no OCR
+fallback. If extraction exceeds a configured limit, the note records that the input was truncated.
+
+### 5. Build the brief and index
+
+Claude Code reads every detailed note and produces the daily brief. The code verifies that each source-note wikilink is
+present and appends any missing links before writing the file. It then rebuilds `daily/YYYY-MM-DD.md` from that day's
+Markdown frontmatter.
+
+### 6. Optionally notify DingTalk
+
+The final step sends the brief body, without YAML frontmatter, to each configured DingTalk group in order. With no
+conversation IDs it is a no-op. If one group fails, the step still attempts the remaining groups and reports the
+combined failure afterward.
+
+## Dates, reruns, and idempotency
+
+- `date` must be an exact `YYYY-MM-DD` value. When omitted, the job uses today in the application timezone, which is
+  `Asia/Shanghai` in the built-in configuration.
+- “Yesterday” means `date - 1 day`, not the previous 24 hours.
+- `history_days` considers prior dated note directories only; the current run date is never part of its history scan.
+- If `daily/<date>/daily-paper-brief.md` already exists and `force=false`, collection, ranking, model calls, PDF work,
+  and digest generation are skipped. The existing brief remains available to the DingTalk notification step.
+- `force=true` regenerates the notes and brief. Existing valid PDFs are still reused.
+
+Each PDF, detailed note, and final brief uses a temporary file followed by replacement so callers do not see a
+partially written file. The complete multi-file workflow is not transactional, and there is no global lock for two
+concurrent runs of the same date.
+
+## Running the cookbook
+
+The standalone configuration defines three jobs:
+
+| Job                | Behavior                                                |
+|--------------------|---------------------------------------------------------|
+| `daily_paper`      | On-demand generation through the CLI or HTTP service    |
+| `daily_paper_cron` | The same pipeline every day at 08:00 in `Asia/Shanghai` |
+| `dingtalk_wait`    | A supervised background DingTalk agent                  |
+
+### One-time runs
+
+Generate today's brief:
+
+```bash
+reme start config=daily_cookbook job=daily_paper
+```
+
+Generate a specific date with selected overrides:
+
+```bash
+reme start \
+  config=daily_cookbook \
+  job=daily_paper \
+  date=2026-07-21 \
+  top_k=3 \
+  history_days=30
+```
+
+Regenerate a date whose brief already exists:
+
+```bash
+reme start config=daily_cookbook job=daily_paper date=2026-07-21 force=true
+```
+
+Add `service.show_metadata=true` to a one-time command when the response metadata is useful for diagnostics.
+
+### Long-running service and cron
+
+Start the standalone HTTP service and its scheduled/background jobs:
+
+```bash
+reme start config=daily_cookbook
+```
+
+It listens on `127.0.0.1:8001` by default, so it can run beside the default ReMe service. Call the on-demand job from
+another terminal with either the ReMe client or HTTP:
+
+```bash
+reme daily_paper host=127.0.0.1 port=8001
+```
+
+```bash
+curl -s http://127.0.0.1:8001/daily_paper \
+  -H 'Content-Type: application/json' \
+  -d '{"date":"2026-07-21","top_k":3,"force":false}'
+```
+
+Service and schedule settings can be overridden at startup:
+
+```bash
+reme start \
+  config=daily_cookbook \
+  service.host=0.0.0.0 \
+  service.port=8101 \
+  jobs.daily_paper_cron.cron="30 7 * * *"
+```
+
+## Configuration
+
+The most useful job settings are:
+
+| Setting           |      Default | Purpose                                                      |
+|-------------------|-------------:|--------------------------------------------------------------|
+| `candidate_limit` |         `20` | Maximum number of papers sent to selection                   |
+| `memory_reserve`  |          `5` | Candidate positions reserved by the memory-keyword heuristic |
+| `top_k`           |          `3` | Number of papers selected and analyzed                       |
+| `rrf_k`           |         `60` | Reciprocal-rank fusion constant                              |
+| `weekly_weight`   |        `0.7` | Weight of the weekly ranking in fusion                       |
+| `history_days`    |         `30` | Prior recommendation window excluded by arXiv ID             |
+| `hf_timeout`      | `30` seconds | Hugging Face request timeout                                 |
+| `hf_max_retries`  |          `3` | Maximum Hugging Face request attempts                        |
+| `pdf_timeout`     | `90` seconds | arXiv download timeout                                       |
+| `max_pdf_bytes`   |   `52428800` | Maximum PDF size (50 MiB)                                    |
+| `max_pdf_pages`   |         `80` | Maximum pages extracted for analysis                         |
+| `max_pdf_chars`   |     `240000` | Maximum extracted characters sent for one paper              |
+
+The public job parameters are `date`, `force`, `top_k`, `weekly_weight`, and `history_days`. Explicit invocation values
+take precedence over the job defaults.
+
+The standalone application also accepts these environment variables:
+
+| Variable                                | Purpose                                         |
+|-----------------------------------------|-------------------------------------------------|
+| `DAILY_PAPER_WORKSPACE_DIR`             | Overrides the default `.reme` workspace         |
+| `DAILY_PAPER_PROJECT_PATH`              | Repository/project path visible to Claude Code  |
+| `DAILY_PAPER_HOST` / `DAILY_PAPER_PORT` | HTTP bind address                               |
+| `CLAUDE_CODE_API_KEY`                   | API key for the configured Claude Code endpoint |
+| `CLAUDE_CODE_MODEL_NAME`                | Model override                                  |
+| `CLAUDE_CODE_BASE_URL`                  | Anthropic-compatible endpoint override          |
+
+`DAILY_PAPER_PROJECT_PATH` defaults to `..` relative to the workspace. With the default `.reme` workspace, starting
+from the repository root resolves it back to the repository. If the workspace lives elsewhere, set both paths
+explicitly.
+
+ReMe loads an uncommitted `.env` file found from the current directory upward, so the same values may be placed there
+instead of exported in the shell.
+
+## DingTalk configuration
+
+DingTalk is optional. Configure it only when brief delivery or the background DingTalk agent is needed:
+
+```dotenv
+DINGTALK_APP_KEY=your-app-key
+DINGTALK_APP_SECRET=your-app-secret
+DINGTALK_ROBOT_CODE=your-robot-code
+DINGTALK_CONVERSATION_IDS=cid-group-one,cid-group-two
+```
+
+`DINGTALK_CONVERSATION_IDS` is required only for proactive brief delivery. The background `dingtalk_wait` job uses the
+first three credentials but not the conversation list.
+
+## Failure recovery and boundaries
+
+| Situation                            | Behavior                                                             |
+|--------------------------------------|----------------------------------------------------------------------|
+| Temporary Hugging Face failure       | Retries with exponential delay up to `hf_max_retries` attempts       |
+| No eligible papers                   | Fails before ranking                                                 |
+| Invalid `top_k` or selection output  | Fails after validation; selection output gets one retry              |
+| Oversized, invalid, or textless PDF  | Stops during analysis                                                |
+| PDF exceeds page or character limits | Continues with truncated text and records the truncation             |
+| One paper analysis fails             | Stops the job; earlier PDFs and notes remain on disk                 |
+| Brief misses a source-note link      | Appends the missing wikilink before writing                          |
+| Existing final brief                 | Skips generation unless `force=true`; notification can still send it |
+| Concurrent runs for one date         | No pipeline-level lock; later writes may replace earlier results     |
+
+To recover, inspect the date's notes and PDFs, fix the network, credential, model, or PDF issue, then rerun the same
+date with `force=true`. Valid cached PDFs will be reused.
+
+The built-in Claude Code component runs with `permission_mode: bypassPermissions`. ReMe disables Claude Code's
+`WebSearch` tool, and the analysis/digest prompts constrain what the agent should read, but these steps do not set a
+strict per-call tool allowlist or an operating-system sandbox. Run the cookbook only with a trusted project and
+workspace, and tighten the agent configuration before shared or production use.
+
+## Tests
+
+The focused unit suite mocks Hugging Face, arXiv, Claude Code, and DingTalk boundaries:
+
+```bash
+pytest tests/unit/test_daily_paper.py -v
+```
+
+Real runs access external services and may incur model costs; they should not be used as ordinary unit tests.
--- a/cookbook/daily_paper/README_ZH.md
+++ b/cookbook/daily_paper/README_ZH.md
@ -0,0 +1,282 @@
+# 每日论文 Cookbook
+
+[English](README.md)
+
+每日论文是一个本地优先、文件原生的研究资讯工作流。它从 Hugging Face 周榜和月榜采集论文，排除昨日论文和近期
+已经推荐过的论文，对剩余候选进行排序，再使用 Claude Code 生成中文详细论文笔记和约五分钟可读完的中文简报。
+
+工作流由 [`daily_cookbook.yaml`](../../reme/config/daily_cookbook.yaml) 装配，公共 schema 位于
+[`reme/schema/daily_paper.py`](../../reme/schema/daily_paper.py)，各步骤位于
+[`reme/steps/cookbook/daily_paper/`](../../reme/steps/cookbook/daily_paper/)。
+
+## 快速开始
+
+每日论文要求 Python 3.11 或更高版本、`core` 依赖、可访问 Hugging Face 和 arXiv 的网络，以及所配置
+Claude Code endpoint 的凭据。
+
+在仓库根目录运行：
+
+```bash
+python -m pip install -e ".[dev,core]"
+export CLAUDE_CODE_API_KEY="your-api-key"
+reme start config=daily_cookbook job=daily_paper
+```
+
+内置配置默认通过 DashScope 的 Anthropic 兼容 endpoint 使用 `qwen3.7-max`。如需使用其他兼容模型或服务商，
+请覆盖 `CLAUDE_CODE_MODEL_NAME` 和 `CLAUDE_CODE_BASE_URL`。
+
+默认情况下，产物写入 ReMe 启动目录下的 `.reme/`。
+
+## 文件产物
+
+一次成功运行会在 `workspace_dir` 下写入普通 PDF 和 Markdown 文件：
+
+```text
+.reme/
+├── daily/
+│   ├── YYYY-MM-DD.md
+│   └── YYYY-MM-DD/
+│       ├── daily-paper-brief.md
+│       ├── paper-<arxiv-id>.md
+│       └── ...
+├── resource/
+│   └── papers/
+│       ├── <arxiv-id>.pdf
+│       └── ...
+└── mem_session/
+    └── claude_config/
+```
+
+- `paper-<arxiv-id>.md` 是中文详细论文解读，其 YAML frontmatter 会链接原始 PDF 和论文页面。
+- `daily-paper-brief.md` 是约五分钟可读完的中文简报，并包含每篇入选论文笔记的 wikilink。
+- `daily/YYYY-MM-DD.md` 是从当日 Markdown 文件重建的派生日索引。
+- `resource/papers/` 保存可复用的原始 PDF。
+
+论文笔记是推荐历史的事实来源：后续排重会读取其 frontmatter 中的 `arxiv_id`。日索引属于可重建的派生文件。
+当前工作流不会另外写入运行 manifest。
+
+## 工作流程
+
+```mermaid
+flowchart LR
+    HF[Hugging Face<br/>周榜 + 月榜] --> C[1. Collect]
+    Y[昨日论文] --> C
+    H[近期本地笔记] --> C
+    C --> R[2. Rank]
+    R --> S[3. Select]
+    S --> A[4. Analyze PDFs]
+    A --> D[5. Build digest]
+    D --> N[6. Notify DingTalk]
+    A --> P[PDF + 论文笔记]
+    D --> B[简报 + 日索引]
+```
+
+### 1. 采集与排重
+
+Collect 会获取运行日所在 ISO week 的周榜、所在自然月的月榜，以及严格前一个自然日的 Hugging Face Daily Papers
+ID。周榜和月榜元数据按 arXiv ID 合并，同时保留两个榜单各自的展示排名。
+
+随后，它会在配置的历史窗口内扫描 `daily/<prior-date>/paper-*.md`，排除笔记 frontmatter 中已有的 ID。如果排重后
+没有任何可选论文，Job 会明确失败。
+
+### 2. 候选排序
+
+Rank 使用 reciprocal-rank fusion（RRF）：
+
+```text
+score = 1 / (rrf_k + monthly_rank)
+      + weekly_weight / (rrf_k + weekly_rank)
+```
+
+论文缺少某个榜单排名时，该项贡献为零。候选按融合分、upvotes 和 arXiv ID 排序。有界候选池还会为标题或摘要命中
+Agent memory、memory retrieval、continual learning、context compression、knowledge graph、RAG 等记忆相关关键词的
+论文保留若干位置。这个保留策略只是关键词启发式，不是语义分类器。
+
+### 3. 精选论文
+
+Claude Code 接收有界候选池，并返回结构化的 `PaperSelection`。实现要求恰好选择 `top_k` 个候选池内的唯一 ID，
+且 rank 必须连续。输出不合法时，校验错误会反馈给 Agent 并重试一次；第二次仍不合法则 Job 失败。
+
+### 4. 下载并解读 PDF
+
+入选论文按顺序逐篇处理。每篇论文都会经过：
+
+1. 校验当前支持的新版 arXiv ID 格式；
+2. 下载并校验 PDF，或复用文件头为 `%PDF-` 的已有文件；
+3. 使用 `pypdf` 提取文本、插入页码标记，并应用页数和字符数限制；
+4. 请求 Claude Code 返回结构化的详细解读；
+5. 写入规范化 frontmatter 和生成的 Markdown 正文。
+
+当前提取器依赖可用的 PDF 文本层。扫描版或纯图片 PDF 会失败，因为没有 OCR fallback。提取内容超过配置限制时，
+笔记会记录输入已被截断。
+
+### 5. 生成简报与索引
+
+Claude Code 会读取全部详细笔记并生成当日简报。代码会检查每篇源笔记的 wikilink；如有遗漏，会在写入前自动补齐。
+随后，工作流根据当日 Markdown frontmatter 重建 `daily/YYYY-MM-DD.md`。
+
+### 6. 可选的钉钉通知
+
+最后一步会去掉 YAML frontmatter，把简报正文按顺序发送到每个已配置的钉钉群。未配置群会话 ID 时，该步骤无副作用
+跳过。某个群发送失败不会阻止继续尝试其他群，所有发送完成后再汇总报告失败。
+
+## 日期、重跑与幂等
+
+- `date` 必须严格符合 `YYYY-MM-DD`。省略时使用应用配置时区中的今天；内置配置为 `Asia/Shanghai`。
+- “昨日”表示 `date - 1 day`，不是模糊的最近 24 小时。
+- `history_days` 只扫描此前的日期目录，不会把本次运行日纳入历史窗口。
+- 如果 `daily/<date>/daily-paper-brief.md` 已存在且 `force=false`，采集、排序、模型调用、PDF 处理和简报生成都会
+  跳过；已有简报仍会交给钉钉通知步骤。
+- `force=true` 会重新生成笔记和简报，但仍会复用已有且有效的 PDF。
+
+每个 PDF、详细笔记和最终简报都会先写临时文件再替换，避免读取方看到半写状态。整个多文件工作流不是事务，
+同一日期的并发运行也没有全局锁。
+
+## 运行方式
+
+独立配置定义了三个 Job：
+
+| Job                | 行为                                     |
+|--------------------|----------------------------------------|
+| `daily_paper`      | 通过 CLI 或 HTTP 服务按需生成                   |
+| `daily_paper_cron` | 每天 08:00（`Asia/Shanghai`）执行相同 pipeline |
+| `dingtalk_wait`    | 由 supervisor 管理的后台钉钉 Agent             |
+
+### 一次性运行
+
+生成今天的简报：
+
+```bash
+reme start config=daily_cookbook job=daily_paper
+```
+
+生成指定日期，并覆盖部分参数：
+
+```bash
+reme start \
+  config=daily_cookbook \
+  job=daily_paper \
+  date=2026-07-21 \
+  top_k=3 \
+  history_days=30
+```
+
+重新生成已有简报的日期：
+
+```bash
+reme start config=daily_cookbook job=daily_paper date=2026-07-21 force=true
+```
+
+需要查看响应 metadata 进行诊断时，可在一次性命令中加入 `service.show_metadata=true`。
+
+### 常驻服务与 cron
+
+启动独立 HTTP 服务以及定时、后台 Job：
+
+```bash
+reme start config=daily_cookbook
+```
+
+服务默认监听 `127.0.0.1:8001`，因此可以和默认 ReMe 服务并行运行。在另一个终端中通过 ReMe client 或 HTTP
+调用按需任务：
+
+```bash
+reme daily_paper host=127.0.0.1 port=8001
+```
+
+```bash
+curl -s http://127.0.0.1:8001/daily_paper \
+  -H 'Content-Type: application/json' \
+  -d '{"date":"2026-07-21","top_k":3,"force":false}'
+```
+
+监听地址和调度时间可以在启动时覆盖：
+
+```bash
+reme start \
+  config=daily_cookbook \
+  service.host=0.0.0.0 \
+  service.port=8101 \
+  jobs.daily_paper_cron.cron="30 7 * * *"
+```
+
+## 配置
+
+最常用的 Job 配置如下：
+
+| 配置项               |        默认值 | 用途                      |
+|-------------------|-----------:|-------------------------|
+| `candidate_limit` |       `20` | 送入精选阶段的最大论文数            |
+| `memory_reserve`  |        `5` | 记忆关键词启发式保留的候选位置数        |
+| `top_k`           |        `3` | 最终精选和解读的论文数             |
+| `rrf_k`           |       `60` | RRF 常数                  |
+| `weekly_weight`   |      `0.7` | 周榜在融合排序中的权重             |
+| `history_days`    |       `30` | 按 arXiv ID 排除近期推荐的时间窗口  |
+| `hf_timeout`      |     `30` 秒 | Hugging Face 请求 timeout |
+| `hf_max_retries`  |        `3` | Hugging Face 请求最多尝试次数   |
+| `pdf_timeout`     |     `90` 秒 | arXiv 下载 timeout        |
+| `max_pdf_bytes`   | `52428800` | PDF 大小上限（50 MiB）        |
+| `max_pdf_pages`   |       `80` | 最多提取的 PDF 页数            |
+| `max_pdf_chars`   |   `240000` | 单篇论文送入模型的最大提取字符数        |
+
+公开 Job 参数为 `date`、`force`、`top_k`、`weekly_weight` 和 `history_days`。调用时显式传入的值优先于 Job 默认值。
+
+独立应用还支持以下环境变量：
+
+| 环境变量                                    | 用途                                 |
+|-----------------------------------------|------------------------------------|
+| `DAILY_PAPER_WORKSPACE_DIR`             | 覆盖默认 `.reme` workspace             |
+| `DAILY_PAPER_PROJECT_PATH`              | Claude Code 可见的仓库或项目路径             |
+| `DAILY_PAPER_HOST` / `DAILY_PAPER_PORT` | HTTP 监听地址                          |
+| `CLAUDE_CODE_API_KEY`                   | 所配置 Claude Code endpoint 的 API key |
+| `CLAUDE_CODE_MODEL_NAME`                | 覆盖模型名称                             |
+| `CLAUDE_CODE_BASE_URL`                  | 覆盖 Anthropic 兼容 endpoint           |
+
+`DAILY_PAPER_PROJECT_PATH` 默认是相对于 workspace 的 `..`。使用默认 `.reme` workspace 并从仓库根目录启动时，
+它会解析回仓库根目录。如果 workspace 位于其他位置，请显式设置这两个路径。
+
+ReMe 会从当前目录向上查找未提交的 `.env`，因此也可以把相同变量放在其中，而不是在 shell 中导出。
+
+## 钉钉配置
+
+钉钉是可选能力。仅在需要投递简报或运行后台钉钉 Agent 时配置：
+
+```dotenv
+DINGTALK_APP_KEY=your-app-key
+DINGTALK_APP_SECRET=your-app-secret
+DINGTALK_ROBOT_CODE=your-robot-code
+DINGTALK_CONVERSATION_IDS=cid-group-one,cid-group-two
+```
+
+只有主动投递简报需要 `DINGTALK_CONVERSATION_IDS`。后台 `dingtalk_wait` Job 使用前三项凭据，不使用群会话列表。
+
+## 故障恢复与边界
+
+| 场景                | 当前行为                            |
+|-------------------|---------------------------------|
+| Hugging Face 暂时失败 | 按指数间隔重试，最多尝试 `hf_max_retries` 次 |
+| 没有 eligible 论文    | 在排序前失败                          |
+| `top_k` 或精选结果不合法  | 校验后失败；精选结果可重试一次                 |
+| PDF 太大、无效或没有文本层   | 在解读阶段停止                         |
+| PDF 超过页数或字符数限制    | 使用截断文本继续，并记录截断状态                |
+| 某篇论文解读失败          | Job 停止；此前写入的 PDF 和笔记保留          |
+| 简报遗漏源笔记链接         | 写入前自动补齐 wikilink                |
+| 最终简报已存在           | 默认跳过生成，通知仍可发送；`force=true` 可重跑  |
+| 同一日期并发运行          | 没有 pipeline 级锁，后写入结果可能替换先前结果    |
+
+恢复时，先检查该日期已有的笔记和 PDF，修复网络、凭据、模型或 PDF 问题，再使用相同日期和 `force=true` 重跑。
+有效的缓存 PDF 会被复用。
+
+内置 Claude Code 组件使用 `permission_mode: bypassPermissions`。ReMe 会禁用 Claude Code 的 `WebSearch` 工具，
+Analyze 和 Digest prompt 也会限制 Agent 应读取的内容，但这些步骤没有设置严格的逐次调用工具 allowlist，也不是
+操作系统级沙箱。请只在可信的项目和 workspace 中运行；用于共享或生产环境前，应进一步收紧 Agent 配置。
+
+## 测试
+
+聚焦的单元测试会 mock Hugging Face、arXiv、Claude Code 和钉钉边界：
+
+```bash
+pytest tests/unit/test_daily_paper.py -v
+```
+
+真实运行会访问外部服务并可能产生模型费用，不应把它当作普通单元测试执行。
--- a/pyproject.toml
+++ b/pyproject.toml
@ -44,12 +44,14 @@ dependencies = [
 core = [
    "agentscope==2.0.4.post1",
    "claude-agent-sdk>=0.2.123",
+    "dingtalk-stream>=0.24.3",
+    "openai-codex>=0.144.4",
    "faiss-cpu>=1.13.2",
    "jieba>=0.42.1",
    "rjieba>=0.2.1",
    "neo4j>=6.2.0",
    "networkx>=3.4.2",
-    "openai-codex>=0.144.4",
+    "pypdf>=5.0.0",
 ]
 dev = [
    "pre-commit",
--- a/reme/components/agent_wrapper/as_agent_wrapper.py
+++ b/reme/components/agent_wrapper/as_agent_wrapper.py
@ -1,7 +1,10 @@
 """AgentScope backend for the unified agent wrapper."""

+import asyncio
 import json
+import os
 import re
+import subprocess
 import time
 from collections.abc import AsyncGenerator
 from pathlib import Path
@ -39,6 +42,7 @@ from agentscope.state import AgentState
 from agentscope.tool import (
    Bash,
    Edit,
+    ExecResult,
    FunctionTool,
    Glob,
    Grep,
@ -68,22 +72,56 @@ _UUID_RE = re.compile(


 class WorkspaceBackend(LocalBackend):
-    """LocalBackend whose reported cwd is the configured agent workspace.
+    """Local backend pinned to the agent cwd and configured environment.

    Some AgentScope builtin tools use ``backend.getcwd()`` for default search
    paths or safety checks. Pinning it here keeps those operations aligned with
    the cwd passed to Bash. Tools that require absolute file paths still keep
-    their own validation behavior.
+    their own validation behavior. Subprocesses receive the startup environment
+    captured in ``ApplicationConfig.environment`` in addition to the parent
+    process environment.
    """

-    def __init__(self, cwd: str) -> None:
+    def __init__(self, cwd: str, environment: dict[str, str] | None = None) -> None:
        super().__init__()
        self._workspace_cwd = cwd
+        self._environment = {**os.environ, **(environment or {})}

    async def getcwd(self) -> str:
        """Return the configured workspace directory."""
        return self._workspace_cwd

+    async def exec_shell(
+        self,
+        command: list[str],
+        *,
+        cwd: str | None = None,
+        timeout: float | None = None,
+    ) -> ExecResult:
+        """Run a local subprocess with the configured agent environment."""
+        kwargs: dict[str, Any] = {
+            "env": self._environment,
+            "stderr": asyncio.subprocess.PIPE,
+            "stdout": asyncio.subprocess.PIPE,
+        }
+        if cwd is not None:
+            kwargs["cwd"] = cwd
+        if os.name == "nt":
+            kwargs["creationflags"] = getattr(subprocess, "CREATE_NO_WINDOW", 0x08000000)
+
+        try:
+            process = await asyncio.create_subprocess_exec(*command, **kwargs)
+        except (FileNotFoundError, NotADirectoryError, OSError) as exc:
+            return ExecResult(exit_code=127, stdout=b"", stderr=str(exc).encode("utf-8"))
+
+        try:
+            stdout, stderr = await asyncio.wait_for(process.communicate(), timeout=timeout)
+        except asyncio.TimeoutError:
+            process.kill()
+            await process.communicate()
+            return ExecResult(exit_code=-1, stdout=b"", stderr=b"timed out")
+        return ExecResult(exit_code=process.returncode or 0, stdout=stdout, stderr=stderr)
+

 class BypassAnalysisBash(Bash):
    """Bash variant that delegates permission decisions to PermissionEngine.
@ -140,7 +178,7 @@ class AsAgentWrapper(BaseAgentWrapper):
    ) -> list[ToolBase]:
        """Return selected AgentScope built-in tools rooted at ``self.cwd``."""
        cwd = str(self.cwd)
-        backend = WorkspaceBackend(cwd)
+        backend = WorkspaceBackend(cwd, self.subprocess_environment)
        factories = {
            "bash": lambda: BypassAnalysisBash(cwd=cwd, backend=backend),
            "edit": lambda: Edit(backend=backend),
@ -249,13 +287,7 @@ class AsAgentWrapper(BaseAgentWrapper):

    def _resolve_skills(self, skills: list[str] | str | None) -> list[str]:
        """Resolve configured skill names to AgentScope local skill directories."""
-        if skills is None:
-            return []
-        if skills == "all":
-            return [str(self.project_skills_root)]
-        if isinstance(skills, str):
-            skills = [skills]
-        return [str(self.project_skills_root / skill) for skill in skills]
+        return [str(path) for path in self._resolve_project_skills(skills).values()]

    async def _build_agent(self, inputs: Any, **kwargs) -> tuple[Agent, Any]:
        """Build an Agent instance from kwargs. Returns (agent, processed_inputs)."""
--- a/reme/components/agent_wrapper/base_agent_wrapper.py
+++ b/reme/components/agent_wrapper/base_agent_wrapper.py
@ -22,9 +22,15 @@ class BaseAgentWrapper(BaseComponent):
    component_type = ComponentEnum.AGENT_WRAPPER
    SDK_PACKAGE: ClassVar[str | None] = None

-    def __init__(self, cwd: str | Path | None = None, **kwargs) -> None:
+    def __init__(
+        self,
+        cwd: str | Path | None = None,
+        project_path: str | Path | None = None,
+        **kwargs,
+    ) -> None:
        super().__init__(**kwargs)
        self._cwd = cwd
+        self._project_path = project_path
        if self.SDK_PACKAGE:
            try:
                sdk_version = metadata.version(self.SDK_PACKAGE)
@ -36,8 +42,7 @@ class BaseAgentWrapper(BaseComponent):
    def cwd(self) -> Path:
        """Working directory shared by the agent's shell and file tools.

-        Defaults to the project root (the workspace) — the same directory
-        Claude Code has always used. Override via the ``cwd`` init argument;
+        Defaults to the project root. Override via the ``cwd`` init argument;
        a relative value resolves against the workspace root.
        """
        if not self._cwd:
@ -62,14 +67,53 @@ class BaseAgentWrapper(BaseComponent):

    @property
    def project_path(self) -> Path:
-        """Project root that contains shared assets such as skills."""
-        return self.workspace_path
+        """Project root containing shared assets such as skills.
+
+        A relative configured path resolves from the workspace so applications
+        can keep runtime data in a subdirectory such as ``.reme`` while loading
+        project assets from its parent. The workspace remains the default for
+        backward compatibility.
+        """
+        if not self._project_path:
+            return self.workspace_path
+        project_path = Path(self._project_path).expanduser()
+        if not project_path.is_absolute():
+            project_path = self.workspace_path / project_path
+        return project_path.resolve(strict=False)

    @property
    def project_skills_root(self) -> Path:
        """Project-level skills directory shared by agent backends."""
        return self.project_path / "skills"

+    def _resolve_project_skills(self, skills: list[str] | str | None) -> dict[str, Path]:
+        """Resolve selected skill names to validated project directories."""
+        if skills is None:
+            return {}
+        if skills == "all":
+            if not self.project_skills_root.is_dir():
+                raise FileNotFoundError(f"Project skills directory not found: {self.project_skills_root}")
+            names = sorted(
+                path.name
+                for path in self.project_skills_root.iterdir()
+                if path.is_dir() and (path / "SKILL.md").is_file()
+            )
+        else:
+            names = [skills] if isinstance(skills, str) else list(skills)
+            names = list(dict.fromkeys(names))
+
+        sources: dict[str, Path] = {}
+        for name in names:
+            if not name or Path(name).name != name or name in {".", ".."}:
+                raise ValueError(f"Invalid skill name: {name!r}")
+            source = self.project_skills_root / name
+            if not source.is_dir():
+                raise FileNotFoundError(f"Skill directory not found: {source}")
+            if not (source / "SKILL.md").is_file():
+                raise FileNotFoundError(f"Skill '{name}' is missing SKILL.md: {source}")
+            sources[name] = source
+        return sources
+
    @property
    def subprocess_environment(self) -> dict[str, str]:
        """Configured environment variables for child agent processes."""
--- a/reme/components/agent_wrapper/cc_agent_wrapper.py
+++ b/reme/components/agent_wrapper/cc_agent_wrapper.py
@ -44,24 +44,7 @@ class CcAgentWrapper(BaseAgentWrapper):

    def _ensure_claude_skill_dir(self, config_dir: Path, skills: list[str] | str) -> None:
        """Add selected project skills to Claude Code discovery locations."""
-        project_skills = self.project_skills_root
-        if not project_skills.is_dir():
-            return
-
-        if skills == "all":
-            skill_names = sorted(path.name for path in project_skills.iterdir() if path.is_dir())
-        else:
-            skill_names = list(dict.fromkeys(skills))
-
-        for skill_name in skill_names:
-            if not skill_name or Path(skill_name).name != skill_name or skill_name in {".", ".."}:
-                raise ValueError(f"Invalid skill name: {skill_name!r}")
-
-        sources = {
-            skill_name: project_skills / skill_name
-            for skill_name in skill_names
-            if (project_skills / skill_name).is_dir()
-        }
+        sources = self._resolve_project_skills(skills)
        if not sources:
            return

--- a/reme/components/agent_wrapper/codex_agent_wrapper.py
+++ b/reme/components/agent_wrapper/codex_agent_wrapper.py
@ -130,31 +130,13 @@ class CodexAgentWrapper(BaseAgentWrapper):

    def _ensure_skills(self, skills: list[str] | str | None) -> None:
        """Expose selected project skills through Codex's repo-level directory."""
-        if skills is None:
+        sources = self._resolve_project_skills(skills)
+        if not sources:
            return
-        if skills == "all":
-            if not self.project_skills_root.is_dir():
-                raise FileNotFoundError(f"Project skills directory not found: {self.project_skills_root}")
-            names = sorted(
-                path.name
-                for path in self.project_skills_root.iterdir()
-                if path.is_dir() and (path / "SKILL.md").is_file()
-            )
-        else:
-            names = [skills] if isinstance(skills, str) else list(skills)
-            names = list(dict.fromkeys(names))

-        target_root = self.workspace_path / ".agents" / "skills"
+        target_root = self.project_path / ".agents" / "skills"
        target_root.mkdir(parents=True, exist_ok=True)
-        for name in names:
-            if not name or Path(name).name != name or name in {".", ".."}:
-                raise ValueError(f"Invalid skill name: {name!r}")
-            source = self.project_skills_root / name
-            if not source.is_dir():
-                raise FileNotFoundError(f"Skill directory not found: {source}")
-            if not (source / "SKILL.md").is_file():
-                raise FileNotFoundError(f"Skill '{name}' is missing SKILL.md: {source}")
-
+        for name, source in sources.items():
            target = target_root / name
            if target.is_symlink():
                if target.resolve() == source.resolve():
--- a/reme/config/daily_cookbook.yaml
+++ b/reme/config/daily_cookbook.yaml
@ -0,0 +1,109 @@
+app_name: ReMe Daily Cookbook
+workspace_dir: ${DAILY_PAPER_WORKSPACE_DIR:-.reme}
+timezone: Asia/Shanghai
+language: zh
+
+# This is a standalone application config. It intentionally does not inherit
+# default.yaml and listens on a separate port so it can run beside ReMe.
+service:
+  backend: http
+  host: ${DAILY_PAPER_HOST:-127.0.0.1}
+  port: ${DAILY_PAPER_PORT:-8001}
+
+jobs:
+  daily_paper:
+    backend: base
+    description: "Build detailed readings and a five-minute brief from Hugging Face weekly/monthly papers."
+    candidate_limit: &candidate_limit 20
+    memory_reserve: &memory_reserve 5
+    top_k: &top_k 3
+    rrf_k: &rrf_k 60
+    weekly_weight: &weekly_weight 0.7
+    history_days: &history_days 30
+    hf_timeout: &hf_timeout 30
+    hf_max_retries: &hf_max_retries 3
+    pdf_timeout: &pdf_timeout 90
+    max_pdf_bytes: &max_pdf_bytes 52428800
+    max_pdf_pages: &max_pdf_pages 80
+    max_pdf_chars: &max_pdf_chars 240000
+    parameters:
+      type: object
+      properties:
+        date:
+          type: string
+          description: "Run date in YYYY-MM-DD; empty means today in Asia/Shanghai."
+          default: ""
+        force:
+          type: boolean
+          description: "Regenerate even when that day's final brief already exists."
+          default: false
+        top_k:
+          type: integer
+          description: "Number of papers selected by Claude Code."
+          default: 3
+        weekly_weight:
+          type: number
+          description: "Weekly contribution in reciprocal-rank fusion."
+          default: 0.7
+        history_days:
+          type: integer
+          description: "Prior recommendation window excluded by arXiv ID."
+          default: 30
+    steps: &daily_paper_steps
+      - backend: daily_paper_collect_step
+      - backend: daily_paper_rank_step
+      - backend: daily_paper_select_step
+        agent_wrapper: claude_code
+      - backend: daily_paper_analyze_step
+        agent_wrapper: claude_code
+      - backend: daily_paper_digest_step
+        agent_wrapper: claude_code
+      - backend: dingtalk_markdown_send_step
+        input_mapping:
+          daily_paper_digest_path: markdown_path
+        app_key: ${DINGTALK_APP_KEY:-}
+        app_secret: ${DINGTALK_APP_SECRET:-}
+        robot_code: ${DINGTALK_ROBOT_CODE:-}
+        conversation_ids: ${DINGTALK_CONVERSATION_IDS:-}
+        title: ReMe Daily Paper
+        timeout: 15
+
+  daily_paper_cron:
+    backend: cron
+    cron: "0 8 * * *"
+    candidate_limit: *candidate_limit
+    memory_reserve: *memory_reserve
+    top_k: *top_k
+    rrf_k: *rrf_k
+    weekly_weight: *weekly_weight
+    history_days: *history_days
+    hf_timeout: *hf_timeout
+    hf_max_retries: *hf_max_retries
+    pdf_timeout: *pdf_timeout
+    max_pdf_bytes: *max_pdf_bytes
+    max_pdf_pages: *max_pdf_pages
+    max_pdf_chars: *max_pdf_chars
+    steps: *daily_paper_steps
+
+  dingtalk_wait:
+    backend: background
+    supervisor: true
+    close_timeout: 10
+    steps:
+      - backend: dingtalk_wait_step
+        agent_wrapper: claude_code
+        app_key: ${DINGTALK_APP_KEY:-}
+        app_secret: ${DINGTALK_APP_SECRET:-}
+        robot_code: ${DINGTALK_ROBOT_CODE:-}
+        card_update_interval: 1.0
+        worker_count: 4
+
+components:
+  agent_wrapper:
+    claude_code:
+      backend: claude_code
+      project_path: ${DAILY_PAPER_PROJECT_PATH:-..}
+      model: ${CLAUDE_CODE_MODEL_NAME:-qwen3.7-max}
+      api_key: ${CLAUDE_CODE_API_KEY:-}
+      base_url: ${CLAUDE_CODE_BASE_URL:-https://dashscope.aliyuncs.com/apps/anthropic}
+      permission_mode: bypassPermissions
--- a/reme/schema/init.py
+++ b/reme/schema/init.py
@ -1,6 +1,7 @@
 """Schema"""

 from .application_config import ApplicationConfig, ComponentConfig, JobConfig
+from .daily_paper import DailyBriefOutput, PaperInfo, PaperNoteOutput, PaperSelection, SelectedPaper
 from .dream import (
    DreamExtractOutput,
    DreamState,
@ -22,7 +23,7 @@ from .stream_chunk import StreamChunk
 __all__ = [
    "ApplicationConfig",
    "ComponentConfig",
-    "JobConfig",
+    "DailyBriefOutput",
    "DreamExtractOutput",
    "DreamState",
    "DreamTopic",
@ -33,9 +34,14 @@ __all__ = [
    "FileLink",
    "FileNode",
    "IntegrateOutcome",
+    "JobConfig",
+    "PaperInfo",
+    "PaperNoteOutput",
+    "PaperSelection",
    "ProactiveResult",
    "Request",
    "Response",
+    "SelectedPaper",
    "StreamChunk",
    "TopicSelectionOutput",
 ]
--- a/reme/schema/daily_paper.py
+++ b/reme/schema/daily_paper.py
@ -0,0 +1,71 @@
+"""Typed contracts for the daily-paper cookbook workflow."""
+
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+
+class PaperInfo(BaseModel):
+    """Normalized Hugging Face paper metadata plus local ranking fields."""
+
+    arxiv_id: str
+    title: str = ""
+    summary: str = ""
+    authors: list[str] = Field(default_factory=list)
+    published_at: str | None = None
+    submitted_on_daily_at: str | None = None
+    upvotes: int = 0
+    organization: str | None = None
+    github_repo: str | None = None
+    github_stars: int | None = None
+    project_page: str | None = None
+    thumbnail: str | None = None
+    monthly_rank: int | None = None
+    weekly_rank: int | None = None
+    fused_score: float = 0.0
+
+    @property
+    def hf_url(self) -> str:
+        """Return the canonical Hugging Face paper-page URL."""
+        return f"https://huggingface.co/papers/{self.arxiv_id}"
+
+    @property
+    def arxiv_url(self) -> str:
+        """Return the canonical arXiv abstract URL."""
+        return f"https://arxiv.org/abs/{self.arxiv_id}"
+
+    @property
+    def pdf_url(self) -> str:
+        """Return the canonical arXiv PDF download URL."""
+        return f"https://arxiv.org/pdf/{self.arxiv_id}"
+
+
+class SelectedPaper(BaseModel):
+    """One agent-selected paper."""
+
+    arxiv_id: str
+    rank: int
+    reason: str
+    memory_relevance: Literal["high", "medium", "low"]
+
+
+class PaperSelection(BaseModel):
+    """Structured paper selection result."""
+
+    selection_reasoning: str
+    selected: list[SelectedPaper]
+    alternates: list[str] = Field(default_factory=list)
+
+
+class PaperNoteOutput(BaseModel):
+    """Structured Claude Code output for one detailed paper note."""
+
+    description: str
+    body: str
+
+
+class DailyBriefOutput(BaseModel):
+    """Structured Claude Code output for the final five-minute brief."""
+
+    description: str
+    body: str
--- a/reme/steps/init.py
+++ b/reme/steps/init.py
@ -1,12 +1,13 @@
 """steps"""

-from . import benchmark, common, evolve, file_io, index, transfer
+from . import benchmark, common, cookbook, evolve, file_io, index, transfer
 from .base_step import BaseStep

 __all__ = [
    "BaseStep",
    "benchmark",
    "common",
+    "cookbook",
    "evolve",
    "file_io",
    "index",
--- a/reme/steps/cookbook/init.py
+++ b/reme/steps/cookbook/init.py
@ -0,0 +1,5 @@
+"""Optional, end-to-end cookbook workflows."""
+
+from . import daily_paper, dingtalk
+
+__all__ = ["daily_paper", "dingtalk"]
--- a/reme/steps/cookbook/daily_paper/init.py
+++ b/reme/steps/cookbook/daily_paper/init.py
@ -0,0 +1,15 @@
+"""Daily-paper cookbook workflow."""
+
+from .analyze import DailyPaperAnalyzeStep
+from .collect import DailyPaperCollectStep
+from .digest import DailyPaperDigestStep
+from .rank import DailyPaperRankStep
+from .select import DailyPaperSelectStep
+
+__all__ = [
+    "DailyPaperAnalyzeStep",
+    "DailyPaperCollectStep",
+    "DailyPaperDigestStep",
+    "DailyPaperRankStep",
+    "DailyPaperSelectStep",
+]
--- a/reme/steps/cookbook/daily_paper/_common.py
+++ b/reme/steps/cookbook/daily_paper/_common.py
@ -0,0 +1,77 @@
+"""Shared state and file helpers for daily-paper steps."""
+
+import os
+import re
+from pathlib import Path
+from typing import Any, TypeVar
+from uuid import uuid4
+
+import aiofiles
+import frontmatter
+from pydantic import BaseModel
+
+from ...base_step import BaseStep
+from ...file_io._file_io import get_path_lock
+
+_STATE_PREFIX = "daily_paper_"
+_FRONTMATTER_PATTERN = re.compile(r"^---\s*\n.*?\n---\s*\n", re.DOTALL)
+_OutputT = TypeVar("_OutputT", bound=BaseModel)
+
+
+def structured_output(result: dict[str, Any], model: type[_OutputT]) -> _OutputT:
+    """Validate an agent wrapper's structured output."""
+    value = result.get("structured_output")
+    return value if isinstance(value, model) else model.model_validate(value)
+
+
+def strip_frontmatter(body: str) -> str:
+    """Remove one model-generated YAML frontmatter block."""
+    return _FRONTMATTER_PATTERN.sub("", body.strip(), count=1).strip()
+
+
+async def write_atomic(path: Path, content: str | bytes) -> None:
+    """Write through a sibling temporary file under the repository path lock."""
+    path.parent.mkdir(parents=True, exist_ok=True)
+    lock = await get_path_lock(path)
+    async with lock:
+        temp_path = path.with_name(f".{path.name}.{uuid4().hex}.tmp")
+        payload = content.encode("utf-8") if isinstance(content, str) else content
+        try:
+            async with aiofiles.open(temp_path, "wb") as stream:
+                await stream.write(payload)
+            os.replace(temp_path, path)
+        finally:
+            if temp_path.exists():
+                temp_path.unlink()
+
+
+async def write_markdown(path: Path, body: str, metadata: dict[str, Any]) -> None:
+    """Serialize a frontmatter Markdown document atomically."""
+    rendered = frontmatter.dumps(frontmatter.Post(body.strip(), **metadata))
+    await write_atomic(path, rendered if rendered.endswith("\n") else f"{rendered}\n")
+
+
+class DailyPaperStep(BaseStep):
+    """Shared helpers for steps in one daily-paper RuntimeContext."""
+
+    def _skip(self) -> bool:
+        assert self.context is not None
+        return bool(self.context.get(f"{_STATE_PREFIX}skip", False))
+
+    def _value(self, key: str, default: Any) -> Any:
+        assert self.context is not None
+        return self.context.get(key, self.kwargs.get(key, default))
+
+    def _state(self, key: str) -> Any:
+        assert self.context is not None
+        return self.context.get(f"{_STATE_PREFIX}{key}")
+
+    def _set_state(self, key: str, value: Any) -> None:
+        assert self.context is not None
+        self.context[f"{_STATE_PREFIX}{key}"] = value
+
+    def _run_day(self) -> str:
+        value = self._state("run_date")
+        if not value:
+            raise RuntimeError("daily-paper run date is not initialized")
+        return str(value)
--- a/reme/steps/cookbook/daily_paper/analyze.py
+++ b/reme/steps/cookbook/daily_paper/analyze.py
@ -0,0 +1,140 @@
+"""Download and analyze selected daily-paper PDFs."""
+
+import asyncio
+import datetime as dt
+import json
+from pathlib import Path
+
+from ....components import R
+from ....schema import PaperInfo, PaperNoteOutput, PaperSelection, SelectedPaper
+from ....utils.arxiv import ArxivPdfClient
+from ._common import DailyPaperStep, strip_frontmatter, structured_output, write_markdown
+
+
+@R.register("daily_paper_analyze_step")
+class DailyPaperAnalyzeStep(DailyPaperStep):
+    """Download each selected PDF and use Claude Code for detailed reading."""
+
+    @staticmethod
+    def _extract_pdf_text_sync(path: Path, max_pages: int, max_chars: int) -> tuple[str, int, bool]:
+        try:
+            from pypdf import PdfReader
+        except ImportError as exc:  # pragma: no cover - dependency error has an explicit message
+            raise RuntimeError("pypdf is required for the daily-paper workflow") from exc
+
+        reader = PdfReader(str(path))
+        chunks: list[str] = []
+        size = 0
+        page_count = min(len(reader.pages), max_pages)
+        truncated = len(reader.pages) > max_pages
+        for page_number, page in enumerate(reader.pages[:page_count], start=1):
+            block = f"\n\n--- PAGE {page_number} ---\n\n{(page.extract_text() or '').strip()}"
+            if size + len(block) > max_chars:
+                if (remaining := max_chars - size) > 0:
+                    chunks.append(block[:remaining])
+                truncated = True
+                break
+            chunks.append(block)
+            size += len(block)
+        content = "".join(chunks).strip()
+        if not content:
+            raise ValueError(f"No extractable text found in PDF: {path.name}")
+        return content, len(reader.pages), truncated
+
+    async def _analyze_one(self, paper: PaperInfo, selected: SelectedPaper) -> tuple[str, str]:
+        if self.agent_wrapper is None:
+            raise RuntimeError("Claude Code agent_wrapper is required for paper analysis")
+        day = self._run_day()
+        daily_dir, resource_dir = (
+            str(self.config_value("daily_dir")).strip("/"),
+            str(self.config_value("resource_dir")).strip("/"),
+        )
+        pdf_rel, note_rel = (
+            f"{resource_dir}/papers/{paper.arxiv_id}.pdf",
+            f"{daily_dir}/{day}/paper-{paper.arxiv_id}.md",
+        )
+        pdf_path, note_path = self.workspace_path / pdf_rel, self.workspace_path / note_rel
+        self.logger.info(f"[{self.name}] paper start arxiv_id={paper.arxiv_id}")
+
+        downloader = ArxivPdfClient(
+            timeout=float(self._value("pdf_timeout", 90.0)),
+            max_bytes=int(self._value("max_pdf_bytes", 50 * 1024 * 1024)),
+        )
+        await downloader.download(paper.arxiv_id, pdf_path)
+        self.logger.info(f"[{self.name}] pdf ready arxiv_id={paper.arxiv_id} path={pdf_rel}")
+        pdf_text, page_count, truncated = await asyncio.to_thread(
+            self._extract_pdf_text_sync,
+            pdf_path,
+            int(self._value("max_pdf_pages", 80)),
+            int(self._value("max_pdf_chars", 240_000)),
+        )
+        self.logger.info(
+            f"[{self.name}] pdf extracted arxiv_id={paper.arxiv_id} pages={page_count} "
+            f"chars={len(pdf_text)} truncated={truncated}",
+        )
+        self.logger.info(f"[{self.name}] agent start arxiv_id={paper.arxiv_id}")
+        result = await self.agent_wrapper.reply(
+            self.prompt_format(
+                "analyze_user",
+                paper_info=json.dumps(paper.model_dump(), ensure_ascii=False, indent=2),
+                selection_reason=selected.reason,
+                memory_relevance=selected.memory_relevance,
+                page_count=page_count,
+                truncated=str(truncated).lower(),
+                pdf_text=pdf_text,
+            ),
+            output_schema=PaperNoteOutput,
+        )
+        self.logger.info(f"[{self.name}] agent done arxiv_id={paper.arxiv_id}")
+        output = structured_output(result, PaperNoteOutput)
+        body = strip_frontmatter(output.body)
+        if not output.description.strip() or not body:
+            raise ValueError(f"Claude Code returned an empty paper note for {paper.arxiv_id}")
+        await write_markdown(
+            note_path,
+            body,
+            {
+                "name": f"paper-{paper.arxiv_id}",
+                "description": output.description.strip(),
+                "arxiv_id": paper.arxiv_id,
+                "title": paper.title,
+                "authors": paper.authors,
+                "hf_url": paper.hf_url,
+                "arxiv_url": paper.arxiv_url,
+                "download_url": paper.pdf_url,
+                "source_pdf": f"[[{pdf_rel}]]",
+                "published_at": paper.published_at,
+                "monthly_rank": paper.monthly_rank,
+                "weekly_rank": paper.weekly_rank,
+                "fused_score": round(paper.fused_score, 8),
+                "selection_reason": selected.reason,
+                "memory_relevance": selected.memory_relevance,
+                "generated_at": dt.datetime.now(dt.timezone.utc).isoformat(),
+                "pdf_pages": page_count,
+                "pdf_text_truncated": truncated,
+            },
+        )
+        self.logger.info(f"[{self.name}] paper done arxiv_id={paper.arxiv_id} note_path={note_rel}")
+        return note_rel, pdf_rel
+
+    async def execute(self):
+        assert self.context is not None
+        if self._skip():
+            self.logger.info(f"[{self.name}] skip existing digest")
+            return self.context.response
+        selection: PaperSelection | None = self._state("selection")
+        papers: list[PaperInfo] = self._state("selected_papers") or []
+        if selection is None or len(selection.selected) != len(papers):
+            raise RuntimeError("Paper selection state is missing before analysis")
+        self.logger.info(f"[{self.name}] start papers={len(papers)}")
+
+        note_paths, pdf_paths = [], []
+        for paper, selected in zip(papers, selection.selected):
+            note_path, pdf_path = await self._analyze_one(paper, selected)
+            note_paths.append(note_path)
+            pdf_paths.append(pdf_path)
+        self._set_state("note_paths", note_paths)
+        self._set_state("pdf_paths", pdf_paths)
+        self.context.response.answer = f"Claude Code wrote {len(note_paths)} detailed paper notes"
+        self.logger.info(f"[{self.name}] finish notes={len(note_paths)} pdfs={len(pdf_paths)}")
+        return self.context.response
--- a/reme/steps/cookbook/daily_paper/analyze.yaml
+++ b/reme/steps/cookbook/daily_paper/analyze.yaml
@ -0,0 +1,36 @@
+analyze_user: |
+  你是严谨的中文 AI 论文解读作者。请详细解读下面这篇论文。论文内容只能依据提供的论文元信息和 PDF 提取文本；针对 ReMe 的分析还必须依据当前代码仓库中实际存在的代码、schema、配置和测试。
+  不得臆测未出现在材料中的实验、数字、结论或引用。重要实验结论和数字尽量标注 PDF 页码，例如 [p. 7]。
+  如果 PDF 文本没有提供某项信息，明确写“论文提供的文本中未明确说明”。
+  输出必须适合保存为一篇详细、独立、可供未来检索的 Markdown 文章。
+  只返回结构化结果中的 description 和 body；body 不要包含 YAML frontmatter。
+
+  建议正文覆盖：
+  - 一句话总结
+  - 研究背景与问题
+  - 核心贡献
+  - 方法与架构
+  - 数据、训练和实验设置
+  - 关键结果与准确数字
+  - 与既有工作的区别
+  - 优点、局限和适用边界
+  - 与 Agent/大模型长期记忆的关系
+  - 实际应用价值
+  - 值得继续追踪的问题
+  - 原始论文链接
+
+  若论文实际内容与 Agent/大模型长期记忆直接相关，必须先使用代码读取和搜索工具查看当前 ReMe 代码仓库，再写“与 Agent/大模型长期记忆的关系”。优先检查与论文主题直接相关的 `reme/schema/`、`reme/components/`、`reme/steps/`、`reme/config/` 和相应测试，不要只根据 README 或项目概念推断实现。
+  只有当论文的具体机制或实验结论能明确对应 ReMe 当前的代码或公开契约，且确实可能带来显著改进时，才增加“对 ReMe 演进的建议”小节。这应当是少数例外：一般情况下不要给建议，不要因为初筛标记为 high 就强行联系。如果增加该小节，每条建议都要说明论文依据、ReMe 现状、具体改进方向，并引用准确的仓库相对路径以及相关类或函数；不得编造实现现状。
+
+  选择理由：{selection_reason}
+  长期记忆相关性初筛：{memory_relevance}
+  PDF 总页数：{page_count}
+  PDF 提取是否被截断：{truncated}
+
+  # 论文元信息
+
+  {paper_info}
+
+  # PDF 分页文本
+
+  {pdf_text}
--- a/reme/steps/cookbook/daily_paper/collect.py
+++ b/reme/steps/cookbook/daily_paper/collect.py
@ -0,0 +1,201 @@
+"""Collect ranked Hugging Face papers for the daily-paper workflow."""
+
+import asyncio
+import datetime as dt
+from pathlib import Path
+
+import frontmatter
+
+from ....components import R
+from ....schema import PaperInfo, PaperSelection
+from ....utils.arxiv import ARXIV_ID_PATTERN
+from ....utils.huggingface_papers import HuggingFacePapersClient
+from ...evolve import now
+from ._common import DailyPaperStep
+
+
+@R.register("daily_paper_collect_step")
+class DailyPaperCollectStep(DailyPaperStep):
+    """Collect current weekly/monthly rankings and strict-yesterday exclusions."""
+
+    @staticmethod
+    def _strict_date(value: str) -> dt.date:
+        text = str(value or "").strip()
+        try:
+            parsed = dt.date.fromisoformat(text)
+        except ValueError as exc:
+            raise ValueError("date must be YYYY-MM-DD") from exc
+        if parsed.isoformat() != text:
+            raise ValueError("date must be YYYY-MM-DD")
+        return parsed
+
+    @staticmethod
+    def _paper_scope_values(day: dt.date) -> tuple[str, str]:
+        iso = day.isocalendar()
+        return f"{iso.year}-W{iso.week:02d}", day.strftime("%Y-%m")
+
+    @staticmethod
+    def load_historical_arxiv_ids(
+        workspace: Path,
+        run_date: dt.date,
+        history_days: int,
+        daily_dir: str,
+    ) -> set[str]:
+        """Read previously recommended paper ids from prior daily note frontmatter."""
+        if history_days <= 0:
+            return set()
+        earliest, root = run_date - dt.timedelta(days=history_days), workspace / daily_dir
+        if not root.is_dir():
+            return set()
+
+        found: set[str] = set()
+        for day_dir in root.iterdir():
+            if not day_dir.is_dir():
+                continue
+            try:
+                note_date = dt.date.fromisoformat(day_dir.name)
+            except ValueError:
+                continue
+            if not earliest <= note_date < run_date:
+                continue
+            for note_path in day_dir.glob("paper-*.md"):
+                try:
+                    metadata = frontmatter.load(note_path).metadata
+                except (OSError, UnicodeError, ValueError):
+                    continue
+                arxiv_id = str(metadata.get("arxiv_id") or "").strip()
+                if ARXIV_ID_PATTERN.fullmatch(arxiv_id):
+                    found.add(arxiv_id)
+        return found
+
+    @staticmethod
+    def load_saved_selection(digest_path: Path) -> dict | None:
+        """Rebuild the saved selection from the digest and paper-note frontmatter."""
+        try:
+            digest_metadata = frontmatter.load(digest_path).metadata
+        except (OSError, UnicodeError, ValueError):
+            return None
+        arxiv_ids = digest_metadata.get("arxiv_ids")
+        if not isinstance(arxiv_ids, list) or not arxiv_ids:
+            return None
+
+        selected = []
+        for rank, value in enumerate(arxiv_ids, start=1):
+            arxiv_id = str(value or "").strip()
+            if not ARXIV_ID_PATTERN.fullmatch(arxiv_id):
+                return None
+            try:
+                note_metadata = frontmatter.load(digest_path.parent / f"paper-{arxiv_id}.md").metadata
+            except (OSError, UnicodeError, ValueError):
+                return None
+            selected.append(
+                {
+                    "arxiv_id": arxiv_id,
+                    "rank": rank,
+                    "reason": str(note_metadata.get("selection_reason") or "").strip(),
+                    "memory_relevance": note_metadata.get("memory_relevance"),
+                },
+            )
+
+        try:
+            selection = PaperSelection.model_validate(
+                {
+                    "selection_reasoning": str(digest_metadata.get("selection_reasoning") or "").strip(),
+                    "selected": selected,
+                    "alternates": digest_metadata.get("alternate_arxiv_ids") or [],
+                },
+            )
+        except (TypeError, ValueError):
+            return None
+        return selection.model_dump()
+
+    @staticmethod
+    def _merge_paper(existing: PaperInfo | None, incoming: PaperInfo) -> PaperInfo:
+        if existing is None:
+            return incoming.model_copy(deep=True)
+        values = existing.model_dump()
+        for key, value in incoming.model_dump().items():
+            if key == "upvotes":
+                values[key] = max(int(values[key] or 0), int(value or 0))
+            elif values.get(key) in (None, "", []):
+                values[key] = value
+        return PaperInfo.model_validate(values)
+
+    async def execute(self):
+        assert self.context is not None
+        timezone = self.app_context.app_config.timezone if self.app_context is not None else None
+        raw_date = str(self._value("date", "") or "").strip()
+        run_date = self._strict_date(raw_date) if raw_date else now(timezone).date()
+        day = run_date.isoformat()
+        self._set_state("run_date", day)
+
+        daily_dir = str(self.config_value("daily_dir")).strip("/")
+        digest_rel = f"{daily_dir}/{day}/daily-paper-brief.md"
+        force = bool(self._value("force", False))
+        self.logger.info(f"[{self.name}] start date={day} force={force}")
+        digest_path = self.workspace_path / digest_rel
+        if digest_path.is_file() and not force:
+            self._set_state("skip", True)
+            self._set_state("digest_path", digest_rel)
+            self.context.response.success = True
+            self.context.response.answer = f"Skipped: daily paper brief already exists at {digest_rel}"
+            self.context.response.metadata.update({"date": day, "digest_path": digest_rel, "skipped": True})
+            if selection := self.load_saved_selection(digest_path):
+                self.context.response.metadata["selection"] = selection
+            self.logger.info(f"[{self.name}] skip existing digest path={digest_rel}")
+            return self.context.response
+
+        week, month = self._paper_scope_values(run_date)
+        yesterday = (run_date - dt.timedelta(days=1)).isoformat()
+        self.logger.info(
+            f"[{self.name}] fetch start week={week} month={month} yesterday={yesterday}",
+        )
+        async with HuggingFacePapersClient(
+            timeout=float(self._value("hf_timeout", 30.0)),
+            max_retries=int(self._value("hf_max_retries", 3)),
+        ) as client:
+            weekly, monthly, yesterday_ids = await asyncio.gather(
+                client.fetch_scope("week", week),
+                client.fetch_scope("month", month),
+                client.fetch_daily_ids(yesterday),
+            )
+        self.logger.info(
+            f"[{self.name}] fetch done weekly={len(weekly)} monthly={len(monthly)} " f"yesterday={len(yesterday_ids)}",
+        )
+
+        merged: dict[str, PaperInfo] = {}
+        for rank, paper in enumerate(monthly, start=1):
+            merged[paper.arxiv_id] = self._merge_paper(merged.get(paper.arxiv_id), paper)
+            merged[paper.arxiv_id].monthly_rank = rank
+        for rank, paper in enumerate(weekly, start=1):
+            merged[paper.arxiv_id] = self._merge_paper(merged.get(paper.arxiv_id), paper)
+            merged[paper.arxiv_id].weekly_rank = rank
+
+        historical_ids = self.load_historical_arxiv_ids(
+            self.workspace_path,
+            run_date,
+            int(self._value("history_days", 30)),
+            daily_dir,
+        )
+        eligible = {key: paper for key, paper in merged.items() if key not in yesterday_ids | historical_ids}
+        self.logger.info(
+            f"[{self.name}] filter done merged={len(merged)} excluded_yesterday={len(yesterday_ids)} "
+            f"excluded_history={len(historical_ids)} eligible={len(eligible)}",
+        )
+        if not eligible:
+            raise RuntimeError("No eligible papers remain after yesterday and history exclusions")
+
+        for key, value in {
+            "info": eligible,
+            "week": week,
+            "month": month,
+            "yesterday": yesterday,
+            "excluded_yesterday": sorted(yesterday_ids),
+            "excluded_history": sorted(historical_ids),
+            "source_counts": {"weekly": len(weekly), "monthly": len(monthly), "merged": len(merged)},
+        }.items():
+            self._set_state(key, value)
+        self.context.response.success = True
+        self.context.response.answer = f"Collected {len(eligible)} eligible papers"
+        self.logger.info(f"[{self.name}] finish date={day} eligible={len(eligible)}")
+        return self.context.response
--- a/reme/steps/cookbook/daily_paper/digest.py
+++ b/reme/steps/cookbook/daily_paper/digest.py
@ -0,0 +1,95 @@
+"""Build the final daily-paper brief from detailed notes."""
+
+import datetime as dt
+import json
+from types import SimpleNamespace
+
+from ....components import R
+from ....schema import DailyBriefOutput, PaperSelection
+from ...file_io import refresh_day_index
+from ._common import DailyPaperStep, strip_frontmatter, structured_output, write_markdown
+
+
+@R.register("daily_paper_digest_step")
+class DailyPaperDigestStep(DailyPaperStep):
+    """Use Claude Code to read the detailed notes and create the final brief."""
+
+    async def execute(self):
+        assert self.context is not None
+        if self._skip():
+            self.logger.info(f"[{self.name}] skip existing digest")
+            return self.context.response
+        if self.agent_wrapper is None:
+            raise RuntimeError("Claude Code agent_wrapper is required for the daily brief")
+        note_paths: list[str] = self._state("note_paths") or []
+        selection: PaperSelection | None = self._state("selection")
+        if selection is None or not note_paths:
+            raise RuntimeError("Detailed paper notes are missing before digest generation")
+        self.logger.info(f"[{self.name}] start notes={len(note_paths)}")
+
+        absolute_paths = [str((self.workspace_path / path).resolve()) for path in note_paths]
+        wikilinks = [f"[[{path}]]" for path in note_paths]
+        self.logger.info(f"[{self.name}] agent start notes={len(note_paths)}")
+        result = await self.agent_wrapper.reply(
+            self.prompt_format(
+                "digest_user",
+                top_k=len(note_paths),
+                note_paths=json.dumps(absolute_paths, ensure_ascii=False, indent=2),
+                wikilinks=json.dumps(wikilinks, ensure_ascii=False, indent=2),
+            ),
+            output_schema=DailyBriefOutput,
+        )
+        self.logger.info(f"[{self.name}] agent done notes={len(note_paths)}")
+        output = structured_output(result, DailyBriefOutput)
+        body = strip_frontmatter(output.body)
+        if not output.description.strip() or not body:
+            raise ValueError("Claude Code returned an empty daily paper brief")
+        missing_links = [link for link in wikilinks if link not in body]
+        if missing_links:
+            body += "\n\n## 详细文章\n\n" + "\n".join(f"- {link}" for link in missing_links)
+
+        day = self._run_day()
+        daily_dir = str(self.config_value("daily_dir")).strip("/")
+        digest_rel = f"{daily_dir}/{day}/daily-paper-brief.md"
+        selected_ids = [item.arxiv_id for item in selection.selected]
+        await write_markdown(
+            self.workspace_path / digest_rel,
+            body,
+            {
+                "name": "daily-paper-brief",
+                "description": output.description.strip(),
+                "date": day,
+                "arxiv_ids": selected_ids,
+                "selection_reasoning": selection.selection_reasoning,
+                "alternate_arxiv_ids": selection.alternates,
+                "source_notes": wikilinks,
+                "generated_at": dt.datetime.now(dt.timezone.utc).isoformat(),
+            },
+        )
+        self._set_state("digest_path", digest_rel)
+        self.logger.info(f"[{self.name}] digest written path={digest_rel}")
+        self.logger.info(f"[{self.name}] refresh index start date={day} daily_dir={daily_dir}")
+        await refresh_day_index(SimpleNamespace(workspace_path=self.workspace_path), day, daily_dir)
+        self.logger.info(f"[{self.name}] refresh index done date={day}")
+
+        self.context.response.success = True
+        self.context.response.answer = f"Generated daily paper brief: {digest_rel}"
+        self.context.response.metadata.update(
+            {
+                "date": day,
+                "week": self._state("week"),
+                "month": self._state("month"),
+                "selection_reasoning": selection.selection_reasoning,
+                "selected_arxiv_ids": selected_ids,
+                "note_paths": note_paths,
+                "pdf_paths": self._state("pdf_paths"),
+                "digest_path": digest_rel,
+                "source_counts": self._state("source_counts"),
+                "excluded_yesterday_count": len(self._state("excluded_yesterday") or []),
+                "excluded_history_count": len(self._state("excluded_history") or []),
+            },
+        )
+        self.logger.info(
+            f"[{self.name}] finish date={day} papers={len(selected_ids)} digest_path={digest_rel}",
+        )
+        return self.context.response
--- a/reme/steps/cookbook/daily_paper/digest.yaml
+++ b/reme/steps/cookbook/daily_paper/digest.yaml
@ -0,0 +1,19 @@
+digest_user: |
+  你是中文 AI 研究资讯主编。请依次调用 Read 阅读下面 {top_k} 个绝对路径，忠实阅读指定的详细论文 Markdown，再生成一篇普通读者五分钟可以读懂的每日论文速读。
+  不得只根据文件名或标题写作，必须调用 Read 读取每一个指定文件。不要读取或修改其他文件。
+  保留技术准确性，同时解释三篇论文为什么值得关注、它们之间有什么联系，以及哪些内容与大模型长期记忆有关。
+
+  # 必须读取的 Markdown 路径
+
+  {note_paths}
+
+  # 必须原样出现在正文中的 wikilink
+
+  {wikilinks}
+
+  输出要求：
+  - 只返回结构化结果中的 description 和 body；body 不要包含 YAML frontmatter。
+  - 总阅读时长约五分钟。
+  - 包含“今日一句话”“每篇一分钟读懂”“三篇论文之间的联系”“今天最值得关注什么”等部分。
+  - 每篇介绍必须附上对应的完整 wikilink，不能缩写或改写链接。
+  - 对长期记忆相关论文，明确解释它对 Agent 记忆系统设计的启发。
--- a/reme/steps/cookbook/daily_paper/rank.py
+++ b/reme/steps/cookbook/daily_paper/rank.py
@ -0,0 +1,103 @@
+"""Rank collected papers for the daily-paper workflow."""
+
+from ....components import R
+from ....schema import PaperInfo
+from ._common import DailyPaperStep
+
+_MEMORY_KEYWORDS = (
+    "long-term memory",
+    "long term memory",
+    "lifelong memory",
+    "agent memory",
+    "episodic memory",
+    "memory consolidation",
+    "memory retrieval",
+    "self-evolving memory",
+    "continual learning",
+    "context compression",
+    "personalization",
+    "knowledge graph",
+    "retrieval augmented",
+    "rag",
+    "长期记忆",
+    "记忆整合",
+    "记忆检索",
+)
+
+
+def rrf_score(
+    monthly_rank: int | None,
+    weekly_rank: int | None,
+    *,
+    rrf_k: int = 60,
+    weekly_weight: float = 0.7,
+) -> float:
+    """Fuse optional monthly and weekly ranks with reciprocal-rank fusion."""
+    if rrf_k < 0:
+        raise ValueError("rrf_k must be non-negative")
+    monthly_score = 0.0 if monthly_rank is None else 1.0 / (rrf_k + monthly_rank)
+    weekly_score = 0.0 if weekly_rank is None else weekly_weight / (rrf_k + weekly_rank)
+    return monthly_score + weekly_score
+
+
+def memory_keyword_score(paper: PaperInfo) -> int:
+    """Return a lightweight recall score used to reserve memory-related candidates."""
+    text = f"{paper.title}\n{paper.summary}".lower()
+    return sum(keyword in text for keyword in _MEMORY_KEYWORDS)
+
+
+def build_candidate_pool(papers: list[PaperInfo], *, limit: int = 20, memory_reserve: int = 5) -> list[PaperInfo]:
+    """Keep strong general papers while reserving room for memory-related work."""
+    if limit <= 0:
+        raise ValueError("candidate_limit must be positive")
+    ranked = sorted(papers, key=lambda item: (-item.fused_score, -item.upvotes, item.arxiv_id))
+    reserve = min(max(memory_reserve, 0), limit)
+    selected = ranked[: max(0, limit - reserve)]
+    selected_ids = {paper.arxiv_id for paper in selected}
+    memory_candidates = [
+        paper for paper in ranked if paper.arxiv_id not in selected_ids and memory_keyword_score(paper)
+    ]
+    memory_candidates.sort(
+        key=lambda item: (-memory_keyword_score(item), -item.fused_score, -item.upvotes, item.arxiv_id),
+    )
+    selected.extend(memory_candidates[:reserve])
+    selected_ids = {paper.arxiv_id for paper in selected}
+    selected.extend(paper for paper in ranked if paper.arxiv_id not in selected_ids and len(selected) < limit)
+    return selected[:limit]
+
+
+@R.register("daily_paper_rank_step")
+class DailyPaperRankStep(DailyPaperStep):
+    """Apply RRF and produce the bounded selection pool."""
+
+    async def execute(self):
+        assert self.context is not None
+        if self._skip():
+            self.logger.info(f"[{self.name}] skip existing digest")
+            return self.context.response
+        papers_by_id: dict[str, PaperInfo] = self._state("info") or {}
+        rrf_k, weekly_weight = int(self._value("rrf_k", 60)), float(self._value("weekly_weight", 0.7))
+        candidate_limit = int(self._value("candidate_limit", 20))
+        memory_reserve = int(self._value("memory_reserve", 5))
+        self.logger.info(
+            f"[{self.name}] start papers={len(papers_by_id)} rrf_k={rrf_k} weekly_weight={weekly_weight} "
+            f"candidate_limit={candidate_limit} memory_reserve={memory_reserve}",
+        )
+        for paper in papers_by_id.values():
+            paper.fused_score = rrf_score(
+                paper.monthly_rank,
+                paper.weekly_rank,
+                rrf_k=rrf_k,
+                weekly_weight=weekly_weight,
+            )
+        candidates = build_candidate_pool(
+            list(papers_by_id.values()),
+            limit=candidate_limit,
+            memory_reserve=memory_reserve,
+        )
+        if not candidates:
+            raise RuntimeError("RRF produced no paper candidates")
+        self._set_state("candidates", candidates)
+        self.context.response.answer = f"Ranked {len(candidates)} paper candidates with RRF"
+        self.logger.info(f"[{self.name}] finish candidates={len(candidates)}")
+        return self.context.response
--- a/reme/steps/cookbook/daily_paper/select.py
+++ b/reme/steps/cookbook/daily_paper/select.py
@ -0,0 +1,93 @@
+"""Select final daily papers with Claude Code."""
+
+import json
+
+from ....components import R
+from ....schema import PaperInfo, PaperSelection
+from ._common import DailyPaperStep, structured_output
+from .rank import memory_keyword_score
+
+
+@R.register("daily_paper_select_step")
+class DailyPaperSelectStep(DailyPaperStep):
+    """Use Claude Code to select the final papers."""
+
+    @staticmethod
+    def _validate_selection(selection: PaperSelection, candidates: list[PaperInfo], top_k: int) -> PaperSelection:
+        candidate_ids = {paper.arxiv_id for paper in candidates}
+        ordered = sorted(selection.selected, key=lambda item: item.rank)
+        selected_ids = [item.arxiv_id for item in ordered]
+        if len(ordered) != top_k:
+            raise ValueError(f"Agent selected {len(ordered)} papers; expected {top_k}")
+        if len(set(selected_ids)) != top_k or any(key not in candidate_ids for key in selected_ids):
+            raise ValueError("Agent selection contains duplicate or out-of-pool ids")
+        if [item.rank for item in ordered] != list(range(1, top_k + 1)):
+            raise ValueError("Agent selection ranks must be consecutive starting at 1")
+        alternates = [
+            key for key in dict.fromkeys(selection.alternates) if key in candidate_ids and key not in selected_ids
+        ]
+        return selection.model_copy(update={"selected": ordered, "alternates": alternates})
+
+    async def execute(self):
+        assert self.context is not None
+        if self._skip():
+            self.logger.info(f"[{self.name}] skip existing digest")
+            return self.context.response
+        if self.agent_wrapper is None:
+            raise RuntimeError("Claude Code agent_wrapper is required for paper selection")
+        candidates: list[PaperInfo] = self._state("candidates") or []
+        top_k = int(self._value("top_k", 3))
+        if top_k <= 0 or top_k > len(candidates):
+            raise ValueError(f"top_k must be between 1 and {len(candidates)}")
+        self.logger.info(f"[{self.name}] start candidates={len(candidates)} top_k={top_k}")
+
+        candidate_payload = [
+            {
+                "arxiv_id": paper.arxiv_id,
+                "title": paper.title,
+                "summary": paper.summary,
+                "authors": paper.authors,
+                "organization": paper.organization,
+                "upvotes": paper.upvotes,
+                "monthly_rank": paper.monthly_rank,
+                "weekly_rank": paper.weekly_rank,
+                "fused_score": round(paper.fused_score, 8),
+                "github_repo": paper.github_repo,
+                "github_stars": paper.github_stars,
+                "memory_keyword_score": memory_keyword_score(paper),
+            }
+            for paper in candidates
+        ]
+        feedback, selection = "", None
+        for attempt in range(1, 3):
+            self.logger.info(f"[{self.name}] agent start attempt={attempt}/2 candidates={len(candidates)}")
+            result = await self.agent_wrapper.reply(
+                self.prompt_format(
+                    "select_user",
+                    top_k=top_k,
+                    candidates=json.dumps(candidate_payload, ensure_ascii=False, indent=2),
+                    retry_feedback=feedback or "(none)",
+                ),
+                output_schema=PaperSelection,
+            )
+            try:
+                selection = self._validate_selection(structured_output(result, PaperSelection), candidates, top_k)
+                self.logger.info(f"[{self.name}] agent done attempt={attempt}/2 valid=True")
+                break
+            except (ValueError, TypeError) as exc:
+                feedback = str(exc)
+                self.logger.warning(
+                    f"[{self.name}] agent done attempt={attempt}/2 valid=False error={feedback!r}",
+                )
+        if selection is None:
+            raise RuntimeError(f"Claude Code paper selection failed validation: {feedback}")
+
+        candidate_map = {paper.arxiv_id: paper for paper in candidates}
+        selected_papers = [candidate_map[item.arxiv_id] for item in selection.selected]
+        self._set_state("selection", selection)
+        self._set_state("selected_papers", selected_papers)
+        self.context.response.answer = f"Selected {top_k} papers with Claude Code"
+        self.logger.info(
+            f"[{self.name}] finish selected={','.join(item.arxiv_id for item in selection.selected)}",
+        )
+        return self.context.response
--- a/reme/steps/cookbook/daily_paper/select.yaml
+++ b/reme/steps/cookbook/daily_paper/select.yaml
@ -0,0 +1,18 @@
+select_user: |
+  你是 AI 研究论文编辑。请从以下候选中选择恰好 {top_k} 篇最值得深入阅读的论文，并另外给出至多 3 个候补 ID。
+  选择应兼顾研究价值、技术新颖性、潜在影响和可读性。融合分与榜单排名是重要依据，但不是唯一依据。
+  与大模型 Agent 长期记忆、记忆检索、记忆整合、持续学习、个性化和上下文管理直接相关的高质量论文应优先考虑。
+  只能选择输入候选集合中的 arXiv ID，不得编造论文或事实。输出简洁、可核验的选择理由。
+
+  要求：
+  1. selected 的 rank 必须从 1 到 {top_k} 连续排列。
+  2. arxiv_id 必须逐字复制候选数据中的值，不能重复。
+  3. reason 说明具体选择依据，不要只复述标题。
+  4. memory_relevance 评价论文与大模型/Agent 长期记忆的相关程度。
+  5. selection_reasoning 是面向读者的简短决策摘要，不要输出隐含的逐步思维过程。
+
+  上一次校验反馈：{retry_feedback}
+
+  # 候选论文
+
+  {candidates}
--- a/reme/steps/cookbook/dingtalk/init.py
+++ b/reme/steps/cookbook/dingtalk/init.py
@ -0,0 +1,6 @@
+"""DingTalk cookbook integration."""
+
+from .send import DingTalkMarkdownSendStep
+from .wait import DingTalkWaitStep
+
+__all__ = ["DingTalkMarkdownSendStep", "DingTalkWaitStep"]
--- a/reme/steps/cookbook/dingtalk/send.py
+++ b/reme/steps/cookbook/dingtalk/send.py
@ -0,0 +1,123 @@
+"""Send a workspace Markdown file to DingTalk group conversations."""
+
+import asyncio
+import json
+
+import aiofiles
+import frontmatter
+import httpx
+
+from ....components import R
+from ...base_step import BaseStep
+from ...file_io._path import gate_md, resolve_path
+
+_GROUP_SEND_URL = "https://api.dingtalk.com/v1.0/robot/groupMessages/send"
+
+
+def _conversation_ids(value: str) -> list[str]:
+    return [item.strip() for item in value.split(",") if item.strip()]
+
+
+@R.register("dingtalk_markdown_send_step")
+class DingTalkMarkdownSendStep(BaseStep):
+    """Send one Markdown document serially to configured DingTalk groups."""
+
+    def __init__(
+        self,
+        app_key: str = "",
+        app_secret: str = "",
+        robot_code: str = "",
+        conversation_ids: str = "",
+        title: str = "",
+        timeout: float = 15.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.app_key = app_key
+        self.app_secret = app_secret
+        self.robot_code = robot_code
+        self.conversation_ids = _conversation_ids(conversation_ids)
+        self.title = title
+        self.timeout = timeout
+
+    async def execute(self):
+        assert self.context is not None
+        recipients = self.conversation_ids
+        self.context.response.metadata["dingtalk_configured_count"] = len(recipients)
+        self.context.response.metadata["dingtalk_sent_count"] = 0
+        if not recipients:
+            self.logger.info(f"[{self.name}] skipped DingTalk Markdown delivery: no conversation IDs")
+            return self.context.response
+        if not all((self.app_key, self.app_secret, self.robot_code)):
+            raise RuntimeError("DingTalk Markdown delivery requires app_key, app_secret, and robot_code")
+
+        raw_path = str(self.context.get("markdown_path") or "")
+        if not raw_path:
+            if self.context.response.metadata.get("skipped"):
+                self.logger.info(f"[{self.name}] skipped DingTalk Markdown delivery: no markdown path")
+                return self.context.response
+            raise RuntimeError("DingTalk Markdown delivery requires markdown_path")
+
+        target, error = resolve_path(self.workspace_path, raw_path)
+        if error:
+            raise ValueError(f"Invalid DingTalk Markdown path: {error}")
+        assert target is not None
+        target, is_markdown = gate_md(target)
+        if not is_markdown:
+            raise ValueError("DingTalk Markdown delivery requires a .md file")
+        if not target.is_file():
+            raise FileNotFoundError(f"DingTalk Markdown file does not exist: {raw_path}")
+
+        async with aiofiles.open(target, encoding="utf-8") as stream:
+            document = frontmatter.loads(await stream.read())
+        markdown = document.content.strip()
+        if not markdown:
+            raise ValueError(f"DingTalk Markdown file is empty: {raw_path}")
+
+        import dingtalk_stream  # pylint: disable=import-outside-toplevel
+
+        title = self.title or str(document.metadata.get("name") or target.stem)
+        token_client = dingtalk_stream.DingTalkStreamClient(
+            dingtalk_stream.Credential(self.app_key, self.app_secret),
+        )
+        access_token = await asyncio.to_thread(token_client.get_access_token)
+        if not access_token:
+            raise RuntimeError("Failed to obtain DingTalk access token")
+
+        self.logger.info(
+            f"[{self.name}] sending DingTalk Markdown path={raw_path} recipients={len(recipients)} "
+            f"chars={len(markdown)} timeout={self.timeout:.1f}s",
+        )
+        failures: list[str] = []
+        headers = {"x-acs-dingtalk-access-token": access_token, "User-Agent": "ReMe DingTalk notifier"}
+        transport = httpx.AsyncHTTPTransport(local_address="0.0.0.0")
+        async with httpx.AsyncClient(timeout=self.timeout, headers=headers, transport=transport) as client:
+            for index, conversation_id in enumerate(recipients, start=1):
+                payload = {
+                    "robotCode": self.robot_code,
+                    "openConversationId": conversation_id,
+                    "msgKey": "sampleMarkdown",
+                    "msgParam": json.dumps({"title": title, "text": markdown}, ensure_ascii=False),
+                }
+                try:
+                    response = await client.post(_GROUP_SEND_URL, json=payload)
+                    response.raise_for_status()
+                    result = response.json()
+                    if not isinstance(result, dict) or not result.get("processQueryKey"):
+                        raise ValueError("missing processQueryKey")
+                except (httpx.HTTPError, ValueError) as exc:
+                    failures.append(f"recipient {index}: {type(exc).__name__}")
+                    self.logger.warning(
+                        f"[{self.name}] DingTalk delivery failed recipient={index}/{len(recipients)} "
+                        f"error_type={type(exc).__name__}",
+                    )
+                    continue
+                self.context.response.metadata["dingtalk_sent_count"] += 1
+                self.logger.info(f"[{self.name}] delivered DingTalk Markdown recipient={index}/{len(recipients)}")
+
+        sent_count = self.context.response.metadata["dingtalk_sent_count"]
+        if failures:
+            self.context.response.metadata["dingtalk_delivery_errors"] = failures
+            raise RuntimeError(f"DingTalk Markdown delivery failed for {len(failures)} of {len(recipients)} recipients")
+        self.logger.info(f"[{self.name}] DingTalk Markdown delivery complete sent={sent_count} total={len(recipients)}")
+        return self.context.response
--- a/reme/steps/cookbook/dingtalk/wait.py
+++ b/reme/steps/cookbook/dingtalk/wait.py
@ -0,0 +1,426 @@
+"""Long-running DingTalk Stream bridge for the cookbook application."""
+
+import asyncio
+import contextlib
+import hashlib
+import json
+import time
+from dataclasses import dataclass, field
+from typing import Any
+from urllib.parse import quote_plus
+
+from ...base_step import BaseStep
+from ....components import R
+from ....enumeration import ChunkEnum
+
+_VISIBLE_CHUNKS = {ChunkEnum.THINK, ChunkEnum.TOOL_CALL, ChunkEnum.TOOL_RESULT, ChunkEnum.CONTENT, ChunkEnum.ERROR}
+_CODE_CHUNKS = {ChunkEnum.TOOL_CALL, ChunkEnum.TOOL_RESULT}
+_BLOCK_CHAR_LIMIT = 100
+
+
+def _session_key(message: Any) -> str:
+    """Return the per-sender, per-conversation Claude session key."""
+    parts = (
+        message.conversation_type,
+        message.conversation_id,
+        message.sender_staff_id,
+    )
+    if not all(parts):
+        raise ValueError("DingTalk message requires conversationType, conversationId, and senderStaffId")
+    return ":".join(parts)
+
+
+def _payload_text(value: Any) -> str:
+    text = value if isinstance(value, str) else json.dumps(value, ensure_ascii=False, indent=2)
+    return text.replace("```", "'''")
+
+
+def _session_ref(key: str) -> str:
+    """Return a stable log correlation id without exposing DingTalk identifiers."""
+    return hashlib.sha256(key.encode("utf-8")).hexdigest()[:12]
+
+
+@dataclass
+class _Block:
+    chunk_type: ChunkEnum
+    key: str
+    name: str = ""
+    parts: list[Any] = field(default_factory=list)
+    chars: int = 0
+    truncated: bool = False
+
+
+class _CardRenderer:
+    """Render append-only, size-limited DingTalk blocks."""
+
+    def __init__(self):
+        self.started_at = time.monotonic()
+        self.active: _Block | None = None
+        self.has_output = False
+        self.error = False
+
+    def feed(self, chunk) -> str:
+        """Buffer one block and return completed Markdown blocks."""
+        if chunk.chunk_type not in _VISIBLE_CHUNKS:
+            return ""
+        if chunk.chunk_type == ChunkEnum.ERROR:
+            self.error = True
+
+        key = chunk.block_id or chunk.tool_call_id or chunk.chunk_type.value
+        current = (chunk.chunk_type, key)
+        if not chunk.chunk:
+            return self._complete_active() if self._active_key() == current else ""
+
+        delta = ""
+        if self._active_key() != current:
+            delta += self._complete_active()
+            self.active = _Block(chunk.chunk_type, key, chunk.tool_call_name or "")
+        assert self.active is not None
+        self.has_output = True
+        if chunk.chunk_type in _CODE_CHUNKS:
+            self.active.parts.append(chunk.chunk)
+            return delta
+
+        if self.active.chars == 0:
+            delta += self._text_title(chunk.chunk_type)
+        text = _payload_text(chunk.chunk)
+        if chunk.chunk_type == ChunkEnum.CONTENT:
+            self.active.chars += len(text)
+            return delta + text
+        remaining = _BLOCK_CHAR_LIMIT - self.active.chars
+        delta += text[:remaining]
+        self.active.chars += min(len(text), remaining)
+        if len(text) > remaining and not self.active.truncated:
+            delta += "..."
+            self.active.truncated = True
+        return delta
+
+    def fail(self, message: str) -> str:
+        """Return an error block for an unexpected local failure."""
+        delta = self._complete_active()
+        self.error = True
+        self.has_output = True
+        text = _payload_text(message)
+        body = text if len(text) <= _BLOCK_CHAR_LIMIT else f"{text[:_BLOCK_CHAR_LIMIT]}..."
+        return f"{delta}### ⚠️ Error\n\n{body}\n\n"
+
+    def _active_key(self) -> tuple[ChunkEnum, str] | None:
+        if self.active is None:
+            return None
+        return self.active.chunk_type, self.active.key
+
+    def _complete_active(self) -> str:
+        if self.active is None:
+            return ""
+        block, self.active = self.active, None
+        if block.chunk_type not in _CODE_CHUNKS:
+            return "\n\n"
+        if block.chunk_type == ChunkEnum.TOOL_CALL and len(block.parts) == 1 and self._is_call_metadata(block.parts[0]):
+            return ""
+
+        body = self._block_body(block)
+        body = body if len(body) <= _BLOCK_CHAR_LIMIT else f"{body[:_BLOCK_CHAR_LIMIT]}..."
+        if block.chunk_type == ChunkEnum.TOOL_CALL:
+            name = block.name or self._metadata_name(block) or "tool"
+            safe_name = name.replace("`", "'")
+            title = f"### 🔧 Tool Call · `{safe_name}`"
+        else:
+            title = "### 📦 Tool Result"
+
+        body = "\n".join(f"> {self._visible_indent(line)}" for line in body.splitlines())
+        return f"{title}\n\n{body}\n\n"
+
+    @staticmethod
+    def _text_title(chunk_type: ChunkEnum) -> str:
+        if chunk_type == ChunkEnum.THINK:
+            return "### 🧠 Think\n\n"
+        if chunk_type == ChunkEnum.ERROR:
+            return "### ⚠️ Error\n\n"
+        return "### 💬 Content\n\n"
+
+    def _block_body(self, block: _Block) -> str:
+        if block.chunk_type not in _CODE_CHUNKS:
+            return "".join(_payload_text(part) for part in block.parts)
+        payload, is_json = self._tool_payload(block)
+        if is_json:
+            return json.dumps(payload, ensure_ascii=False, indent=2)
+        return "".join(_payload_text(part) for part in block.parts)
+
+    @staticmethod
+    def _visible_indent(line: str) -> str:
+        """Keep JSON indentation after DingTalk collapses ordinary spaces."""
+        spaces = len(line) - len(line.lstrip(" "))
+        return "　" * (spaces // 2) + line[spaces:]
+
+    def _tool_payload(self, block: _Block) -> tuple[Any, bool]:
+        parts = block.parts
+        if block.chunk_type == ChunkEnum.TOOL_CALL and len(parts) > 1 and self._is_call_metadata(parts[0]):
+            parts = parts[1:]
+        if len(parts) == 1 and not isinstance(parts[0], str):
+            return self._expand_nested_json(parts[0]), True
+        if parts and all(isinstance(part, str) for part in parts):
+            try:
+                value = json.loads("".join(parts))
+            except json.JSONDecodeError:
+                pass
+            else:
+                return self._expand_nested_json(value), True
+        return None, False
+
+    def _metadata_name(self, block: _Block) -> str:
+        if not block.parts or not self._is_call_metadata(block.parts[0]):
+            return ""
+        return json.loads(block.parts[0]).get("name") or ""
+
+    @staticmethod
+    def _is_call_metadata(value: Any) -> bool:
+        if not isinstance(value, str):
+            return False
+        try:
+            parsed = json.loads(value)
+        except json.JSONDecodeError:
+            return False
+        return isinstance(parsed, dict) and bool(parsed) and set(parsed) <= {"id", "name"}
+
+    @classmethod
+    def _expand_nested_json(cls, value: Any) -> Any:
+        if isinstance(value, dict):
+            return {key: cls._expand_nested_json(item) for key, item in value.items()}
+        if isinstance(value, list):
+            return [cls._expand_nested_json(item) for item in value]
+        if isinstance(value, str) and value.lstrip().startswith(("{", "[")):
+            try:
+                return cls._expand_nested_json(json.loads(value))
+            except json.JSONDecodeError:
+                pass
+        return value
+
+    def finish(self) -> str:
+        """Close the active block and append elapsed time."""
+        delta = self._complete_active()
+        if not delta and self.has_output:
+            delta = "\n\n"
+        return f"{delta}{time.monotonic() - self.started_at:.1f}s"
+
+
+@R.register("dingtalk_wait_step")
+class DingTalkWaitStep(BaseStep):
+    """Receive DingTalk messages and stream Claude Code replies into AI cards."""
+
+    def __init__(
+        self,
+        app_key: str = "",
+        app_secret: str = "",
+        robot_code: str = "",
+        card_update_interval: float = 1.0,
+        worker_count: int = 4,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.app_key = app_key
+        self.app_secret = app_secret
+        self.robot_code = robot_code
+        self.card_update_interval = max(0.05, card_update_interval)
+        self.worker_count = max(1, worker_count)
+
+    async def execute(self):
+        assert self.context is not None
+        if self.context.stop_event is None or self.app_context is None:
+            raise RuntimeError("dingtalk_wait_step requires an ApplicationContext and background stop_event")
+        if self.agent_wrapper is None:
+            raise RuntimeError("dingtalk_wait_step requires an agent_wrapper")
+        if not self.app_key or not self.app_secret or not self.robot_code:
+            raise RuntimeError("dingtalk_wait_step requires app_key, app_secret, and robot_code")
+
+        import dingtalk_stream  # pylint: disable=import-outside-toplevel
+
+        queue: asyncio.Queue = asyncio.Queue()
+        handler = self._make_handler(dingtalk_stream, queue)
+        client = dingtalk_stream.DingTalkStreamClient(
+            dingtalk_stream.Credential(self.app_key, self.app_secret),
+        )
+        client.register_callback_handler(dingtalk_stream.ChatbotMessage.TOPIC, handler)
+        sessions = self.app_context.metadata.setdefault("dingtalk_agent_sessions", {})
+        locks: dict[str, asyncio.Lock] = {}
+        self.logger.info(
+            f"[{self.name}] starting DingTalk Stream bridge "
+            f"workers={self.worker_count} card_update_interval={self.card_update_interval:.2f}s",
+        )
+        workers = [
+            asyncio.create_task(self._worker(queue, locks, sessions, handler, dingtalk_stream))
+            for _ in range(self.worker_count)
+        ]
+        try:
+            await self._run_client(client, self.context.stop_event)
+        finally:
+            for worker in workers:
+                worker.cancel()
+            await asyncio.gather(*workers, return_exceptions=True)
+            self.logger.info(f"[{self.name}] DingTalk Stream bridge stopped")
+        return self.context.response
+
+    @staticmethod
+    def _make_handler(dingtalk_stream, queue: asyncio.Queue):
+        class QueueHandler(dingtalk_stream.ChatbotHandler):
+            """Acknowledge callbacks after placing them on the worker queue."""
+
+            async def process(self, callback):
+                """Enqueue one callback and immediately acknowledge it."""
+                queue.put_nowait(dingtalk_stream.ChatbotMessage.from_dict(callback.data))
+                return dingtalk_stream.AckMessage.STATUS_OK, "OK"
+
+        return QueueHandler()
+
+    async def _worker(self, queue, locks, sessions, handler, dingtalk_stream) -> None:
+        while True:
+            message = await queue.get()
+            try:
+                key = _session_key(message)
+                async with locks.setdefault(key, asyncio.Lock()):
+                    await self._handle_message(message, key, sessions, handler, dingtalk_stream)
+            except Exception as exc:  # A bad message must not disconnect the Stream client.
+                self.logger.exception(f"Failed to handle DingTalk message: {exc}")
+                await asyncio.to_thread(handler.reply_text, f"处理失败：{exc}", message)
+            finally:
+                queue.task_done()
+
+    async def _handle_message(self, message, key, sessions, handler, dingtalk_stream) -> None:
+        session_ref = _session_ref(key)
+        self.logger.info(
+            f"[{self.name}] handling DingTalk callback session={session_ref} "
+            f"conversation_type={message.conversation_type!r} conversation_id={message.conversation_id!r} "
+            f"sender_staff_id={message.sender_staff_id!r}",
+        )
+        if self.robot_code and getattr(message, "robot_code", "") != self.robot_code:
+            self.logger.warning(
+                f"[{self.name}] rejected DingTalk callback session={session_ref} reason=robot_code_mismatch",
+            )
+            raise ValueError("DingTalk callback robotCode does not match configured robot_code")
+        text = (message.text.content if message.text else "").strip()
+        if not text:
+            self.logger.info(f"[{self.name}] ignored non-text DingTalk message session={session_ref}")
+            await asyncio.to_thread(handler.reply_text, "暂时只支持文本消息。", message)
+            return
+        if text == "/clear":
+            cleared = sessions.pop(key, None) is not None
+            self.logger.info(f"[{self.name}] cleared DingTalk session session={session_ref} existed={cleared}")
+            await asyncio.to_thread(
+                handler.reply_text,
+                "✅ Conversation cleared. The next message will start a new session.",
+                message,
+            )
+            return
+
+        resumed = key in sessions
+        self.logger.info(
+            f"[{self.name}] received DingTalk text session={session_ref} chars={len(text)} resume={resumed}",
+        )
+        renderer = _CardRenderer()
+        card = dingtalk_stream.AIMarkdownCardInstance(handler.dingtalk_client, message)
+        card_id = await card.async_create_and_send_card(
+            card.card_template_id,
+            card.get_card_data(flow_status=dingtalk_stream.AICardStatus.PROCESSING),
+            at_sender=True,
+        )
+        if not card_id:
+            raise RuntimeError("创建钉钉 AI 卡片失败")
+        self.logger.debug(f"[{self.name}] started DingTalk AI card session={session_ref}")
+
+        last_update = 0.0
+        pending = ""
+        chunk_count = 0
+        update_count = 0
+        streamed_chars = 0
+        finalizing = False
+        try:
+            kwargs = {"resume": sessions[key]} if key in sessions else {}
+            async for chunk in self.agent_wrapper.reply_stream(text, **kwargs):
+                chunk_count += 1
+                if chunk.session_id:
+                    sessions[key] = chunk.session_id
+                pending += renderer.feed(chunk)
+                if pending and time.monotonic() - last_update >= self.card_update_interval:
+                    delta = pending
+                    await card.async_streaming(
+                        card_id,
+                        "msgContent",
+                        delta,
+                        append=True,
+                        finished=False,
+                        failed=False,
+                    )
+                    pending = ""
+                    streamed_chars += len(delta)
+                    update_count += 1
+                    last_update = time.monotonic()
+
+            pending += renderer.finish()
+            finalizing = True
+            await card.async_streaming(
+                card_id,
+                "msgContent",
+                pending,
+                append=True,
+                finished=True,
+                failed=renderer.error,
+            )
+            streamed_chars += len(pending)
+            pending = ""
+            log = self.logger.warning if renderer.error else self.logger.info
+            log(
+                f"[{self.name}] completed DingTalk reply session={session_ref} success={not renderer.error} "
+                f"chunks={chunk_count} card_updates={update_count} card_chars={streamed_chars} "
+                f"elapsed={time.monotonic() - renderer.started_at:.2f}s",
+            )
+        except Exception:
+            if not finalizing:
+                if not renderer.error:
+                    pending += renderer.fail("Agent 执行失败")
+                pending += renderer.finish()
+            failure_delta, pending = pending, ""
+            with contextlib.suppress(Exception):
+                await card.async_streaming(
+                    card_id,
+                    "msgContent",
+                    failure_delta,
+                    append=True,
+                    finished=True,
+                    failed=True,
+                )
+            streamed_chars += len(failure_delta)
+            self.logger.warning(
+                f"[{self.name}] DingTalk reply failed session={session_ref} "
+                f"chunks={chunk_count} card_updates={update_count} card_chars={streamed_chars} "
+                f"elapsed={time.monotonic() - renderer.started_at:.2f}s",
+            )
+            raise
+
+    @staticmethod
+    async def _run_client(client, stop_event: asyncio.Event) -> None:
+        """Run one cancellable WebSocket connection; BackgroundJob owns retries."""
+        import websockets  # pylint: disable=import-outside-toplevel
+
+        client.pre_start()
+        connection = await asyncio.to_thread(client.open_connection)
+        if not connection:
+            raise ConnectionError("DingTalk open connection failed")
+        uri = f'{connection["endpoint"]}?ticket={quote_plus(connection["ticket"])}'
+        async with websockets.connect(uri) as websocket:
+            client.websocket = websocket
+            keepalive = asyncio.create_task(client.keepalive(websocket))
+
+            async def close_when_stopped() -> None:
+                await stop_event.wait()
+                await websocket.close()
+
+            stopper = asyncio.create_task(close_when_stopped())
+            try:
+                async for raw_message in websocket:
+                    if await client.route_message(json.loads(raw_message)) == client.TAG_DISCONNECT:
+                        await websocket.close()
+            finally:
+                for task in (stopper, keepalive):
+                    task.cancel()
+                await asyncio.gather(stopper, keepalive, return_exceptions=True)
+            if not stop_event.is_set():
+                raise ConnectionError("DingTalk WebSocket closed")
--- a/reme/utils/arxiv.py
+++ b/reme/utils/arxiv.py
@ -0,0 +1,59 @@
+"""arXiv validation and PDF download helpers."""
+
+import os
+import re
+from pathlib import Path
+from uuid import uuid4
+
+import aiofiles
+import httpx
+
+ARXIV_ID_PATTERN = re.compile(r"^\d{4}\.\d{4,5}$")
+
+
+class ArxivPdfClient:
+    """Download validated arXiv PDFs to local files."""
+
+    def __init__(self, *, timeout: float = 90.0, max_bytes: int = 50 * 1024 * 1024) -> None:
+        self.timeout, self.max_bytes = timeout, max_bytes
+
+    async def download(self, arxiv_id: str, target: Path) -> Path:
+        """Download one PDF atomically, reusing an existing valid target."""
+        if not ARXIV_ID_PATTERN.fullmatch(arxiv_id):
+            raise ValueError(f"Invalid arXiv id: {arxiv_id!r}")
+        if target.is_file() and target.stat().st_size > 5:
+            with target.open("rb") as existing:
+                if existing.read(5) == b"%PDF-":
+                    return target
+
+        target.parent.mkdir(parents=True, exist_ok=True)
+        part_path = target.with_name(f".{target.name}.{uuid4().hex}.part")
+        size = 0
+        try:
+            async with httpx.AsyncClient(
+                timeout=self.timeout,
+                follow_redirects=True,
+                headers={"User-Agent": "ReMe arXiv client"},
+            ) as client:
+                async with client.stream("GET", f"https://arxiv.org/pdf/{arxiv_id}") as response:
+                    response.raise_for_status()
+                    content_length = int(response.headers.get("content-length") or 0)
+                    if content_length and content_length > self.max_bytes:
+                        raise ValueError(f"PDF exceeds maximum size: {content_length} > {self.max_bytes}")
+                    async with aiofiles.open(part_path, "wb") as stream:
+                        async for chunk in response.aiter_bytes():
+                            size += len(chunk)
+                            if size > self.max_bytes:
+                                raise ValueError(f"PDF exceeds maximum size: {size} > {self.max_bytes}")
+                            await stream.write(chunk)
+            if size <= 5:
+                raise ValueError(f"Downloaded PDF is empty for {arxiv_id}")
+            async with aiofiles.open(part_path, "rb") as stream:
+                header = await stream.read(5)
+            if header != b"%PDF-":
+                raise ValueError(f"Downloaded content is not a PDF for {arxiv_id}")
+            os.replace(part_path, target)
+            return target
+        finally:
+            if part_path.exists():
+                part_path.unlink()
--- a/reme/utils/huggingface_papers.py
+++ b/reme/utils/huggingface_papers.py
@ -0,0 +1,156 @@
+"""Client and response normalization for Hugging Face Papers."""
+
+import asyncio
+import re
+from collections.abc import Iterable
+from typing import Any
+
+import httpx
+
+from ..schema import PaperInfo
+from .arxiv import ARXIV_ID_PATTERN
+
+HF_BASE_URL = "https://huggingface.co"
+_PAPER_LINK_PATTERN = re.compile(
+    r"href=[\"'](?:https://huggingface\.co)?/papers/(\d{4}\.\d{4,5})(?:[^\"']*)?[\"']",
+    re.IGNORECASE,
+)
+
+
+def _organization_name(value: Any) -> str | None:
+    if not isinstance(value, dict):
+        return None
+    name = value.get("fullname") or value.get("name")
+    return str(name).strip() if name else None
+
+
+def paper_info_from_payload(item: dict[str, Any]) -> PaperInfo:
+    """Normalize either a daily-list item or a paper-detail response."""
+    nested = item.get("paper")
+    paper = nested if isinstance(nested, dict) else item
+    arxiv_id = str(paper.get("id") or "").strip()
+    if not ARXIV_ID_PATTERN.fullmatch(arxiv_id):
+        raise ValueError(f"Invalid arXiv id from Hugging Face: {arxiv_id!r}")
+
+    authors = [
+        str(author["name"]).strip()
+        for author in paper.get("authors") or []
+        if isinstance(author, dict) and author.get("name")
+    ]
+
+    organization = paper.get("organization") or item.get("organization")
+    return PaperInfo(
+        arxiv_id=arxiv_id,
+        title=str(paper.get("title") or item.get("title") or "").strip(),
+        summary=str(paper.get("summary") or item.get("summary") or "").strip(),
+        authors=authors,
+        published_at=paper.get("publishedAt") or item.get("publishedAt"),
+        submitted_on_daily_at=paper.get("submittedOnDailyAt"),
+        upvotes=int(paper.get("upvotes") or 0),
+        organization=_organization_name(organization),
+        github_repo=paper.get("githubRepo"),
+        github_stars=paper.get("githubStars"),
+        project_page=paper.get("projectPage"),
+        thumbnail=item.get("thumbnail"),
+    )
+
+
+def paper_ids_from_html(content: str) -> list[str]:
+    """Return unique paper ids in server-rendered display order."""
+    return list(dict.fromkeys(_PAPER_LINK_PATTERN.findall(content)))
+
+
+class HuggingFacePapersClient:
+    """Fetch ranked paper pages and their accompanying JSON metadata."""
+
+    def __init__(
+        self,
+        *,
+        client: httpx.AsyncClient | None = None,
+        timeout: float = 30.0,
+        max_retries: int = 3,
+        detail_concurrency: int = 5,
+    ) -> None:
+        self._owns_client = client is None
+        self.client = client or httpx.AsyncClient(
+            base_url=HF_BASE_URL,
+            timeout=timeout,
+            follow_redirects=True,
+            headers={"User-Agent": "ReMe daily-paper cookbook"},
+        )
+        self.max_retries = max(1, int(max_retries))
+        self.detail_concurrency = max(1, int(detail_concurrency))
+
+    async def __aenter__(self) -> "HuggingFacePapersClient":
+        return self
+
+    async def __aexit__(self, *_args) -> None:
+        if self._owns_client:
+            await self.client.aclose()
+
+    async def _get(self, path: str, *, params: dict[str, Any] | None = None) -> httpx.Response:
+        last_error: Exception | None = None
+        for attempt in range(self.max_retries):
+            try:
+                response = await self.client.get(path, params=params)
+                response.raise_for_status()
+                return response
+            except httpx.HTTPError as exc:
+                last_error = exc
+                if attempt + 1 >= self.max_retries:
+                    break
+                await asyncio.sleep(0.25 * (2**attempt))
+        assert last_error is not None
+        raise last_error
+
+    async def _json_list(self, params: dict[str, Any]) -> list[dict[str, Any]]:
+        response = await self._get("/api/daily_papers", params=params)
+        payload = response.json()
+        if not isinstance(payload, list):
+            raise ValueError(f"Unexpected Hugging Face daily_papers response: {type(payload).__name__}")
+        return [item for item in payload if isinstance(item, dict)]
+
+    async def fetch_detail(self, arxiv_id: str) -> PaperInfo:
+        """Fetch and normalize one paper-detail response."""
+        if not ARXIV_ID_PATTERN.fullmatch(arxiv_id):
+            raise ValueError(f"Invalid arXiv id: {arxiv_id!r}")
+        response = await self._get(f"/api/papers/{arxiv_id}")
+        payload = response.json()
+        if not isinstance(payload, dict):
+            raise ValueError(f"Unexpected paper response for {arxiv_id}")
+        return paper_info_from_payload(payload)
+
+    async def fetch_daily_ids(self, day: str) -> set[str]:
+        """Return paper ids for exactly one Hugging Face Daily Papers date."""
+        items = await self._json_list({"date": day, "limit": 100})
+        return {paper_info_from_payload(item).arxiv_id for item in items}
+
+    async def fetch_scope(self, scope: str, value: str) -> list[PaperInfo]:
+        """Fetch all paper cards shown on one weekly or monthly page."""
+        if scope not in {"week", "month"}:
+            raise ValueError("scope must be 'week' or 'month'")
+
+        page_response, api_items = await asyncio.gather(
+            self._get(f"/papers/{scope}/{value}"),
+            self._json_list({scope: value, "limit": 100}),
+        )
+        page_ids = paper_ids_from_html(page_response.text)
+        api_infos = [paper_info_from_payload(item) for item in api_items]
+        api_by_id = {paper.arxiv_id: paper for paper in api_infos}
+        ranked_ids = page_ids or [paper.arxiv_id for paper in api_infos]
+
+        missing_ids = [arxiv_id for arxiv_id in ranked_ids if arxiv_id not in api_by_id]
+        if missing_ids:
+            details = await self._fetch_details_limited(missing_ids)
+            api_by_id.update({paper.arxiv_id: paper for paper in details})
+
+        return [api_by_id[arxiv_id] for arxiv_id in ranked_ids if arxiv_id in api_by_id]
+
+    async def _fetch_details_limited(self, arxiv_ids: Iterable[str]) -> list[PaperInfo]:
+        semaphore = asyncio.Semaphore(self.detail_concurrency)
+
+        async def fetch(arxiv_id: str) -> PaperInfo:
+            async with semaphore:
+                return await self.fetch_detail(arxiv_id)
+
+        return list(await asyncio.gather(*(fetch(arxiv_id) for arxiv_id in arxiv_ids)))
--- a/skills/dingtalk-message/README.md
+++ b/skills/dingtalk-message/README.md
@ -0,0 +1,985 @@
+# 钉钉消息发送技能 (dingtalk-message)
+
+钉钉消息发送 CLI 工具，支持**企业内部机器人**和 **Webhook 自定义机器人**两种接入方式，支持文本、Markdown、链接、ActionCard、FeedCard、文件等多种消息类型。
+
+> 官方文档：https://open.dingtalk.com/document/development/development-robot-overview
+
+## 目录
+
+- [API 接口文档](#api-接口文档)
+  - [企业内部机器人](#一企业内部机器人-api)
+    - [批量发送单聊消息](#1-批量发送人与机器人会话中机器人消息)
+    - [发送群聊消息](#2-发送群聊消息)
+    - [获取 Access Token](#3-获取-access-token)
+    - [上传媒体文件](#4-上传媒体文件)
+  - [Webhook 自定义机器人](#二webhook-自定义机器人-api)
+    - [Webhook 概述](#webhook-概述)
+    - [安全设置](#安全设置)
+    - [Webhook 消息类型](#webhook-消息类型)
+- [CLI 使用指南](#cli-使用指南)
+- [配置说明](#配置说明)
+- [消息类型详解](#消息类型详解)
+- [Markdown 语法限制](#钉钉-markdown-语法限制)
+- [错误码参考](#错误码参考)
+- [注意事项与限制](#注意事项与限制)
+
+---
+
+## API 接口文档
+
+### 一、企业内部机器人 API
+
+### 1. 批量发送人与机器人会话中机器人消息
+
+> 官方文档：https://open.dingtalk.com/document/development/chatbots-send-one-on-one-chat-messages-in-batches
+
+#### 接口概述
+
+调用本接口批量发送人与机器人会话（即人与机器人的单聊）中的机器人消息。通过该接口，企业内部应用的机器人可以向多个用户同时发送单聊消息。
+
+#### 请求方式
+
+- **HTTP 方法**：`POST`
+- **URL**：`https://api.dingtalk.com/v1.0/robot/oToMessages/batchSend`
+
+#### 请求头
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `x-acs-dingtalk-access-token` | String | 是 | 调用服务端接口的授权凭证（access_token） |
+| `Content-Type` | String | 是 | 固定值：`application/json` |
+
+#### 请求体参数（Body）
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `robotCode` | String | 是 | 机器人的编码（robotCode），可在钉钉开发者后台的机器人管理页面获取 |
+| `userIds` | List\<String\> | 是 | 接收消息的用户 userId 列表，最多支持 **100** 个 |
+| `msgKey` | String | 是 | 消息模板 Key，用于指定消息类型，见下方[消息类型 msgKey 映射表](#消息类型-msgkey-映射表) |
+| `msgParam` | String | 是 | 消息模板参数，JSON 字符串格式，内容结构取决于 `msgKey` 的类型 |
+
+#### 消息类型 msgKey 映射表
+
+| 消息类型 | msgKey 值 | 说明 |
+|----------|-----------|------|
+| 文本消息 | `sampleText` | 纯文本消息，支持 @用户 |
+| 图片消息 | `sampleImage` | 需先上传获取 media_id |
+| 语音消息 | `sampleVoice` | 需先上传获取 media_id |
+| 文件消息 | `sampleFile` | 需先上传获取 media_id |
+| 链接消息 | `sampleLink` | 带缩略图的链接卡片 |
+| Markdown 消息 | `sampleMarkdown` | 支持有限 Markdown 语法 |
+| ActionCard（单按钮） | `sampleActionCard` | 单按钮交互卡片 |
+| ActionCard（多按钮） | `sampleMultiActionCard` | 多按钮交互卡片 |
+
+
+#### 各消息类型 msgParam 结构
+
+**文本消息 (`sampleText`)**
+
+```json
+{
+  "content": "消息内容",
+  "atUserIds": ["userId1", "userId2"]   // 可选，@指定用户
+}
+```
+
+**Markdown 消息 (`sampleMarkdown`)**
+
+```json
+{
+  "title": "消息标题",
+  "text": "#### 标题\n> 引用内容\n正文"
+}
+```
+
+**链接消息 (`sampleLink`)**
+
+```json
+{
+  "title": "链接标题",
+  "text": "链接描述",
+  "messageUrl": "https://example.com",
+  "picUrl": "https://example.com/image.png"   // 可选，缩略图
+}
+```
+
+**图片消息 (`sampleImage`)**
+
+```json
+{
+  "mediaId": "@lADPxxxxxxxx",
+  "caption": "图片描述"   // 可选
+}
+```
+
+**文件消息 (`sampleFile`)**
+
+```json
+{
+  "mediaId": "@lADPxxxxxxxx",
+  "fileName": "报告.pdf",
+  "fileSize": "1024",      // 可选，单位：字节
+  "fileType": "pdf"        // 可选
+}
+```
+
+**语音消息 (`sampleVoice`)**
+
+```json
+{
+  "mediaId": "@lADPxxxxxxxx",
+  "duration": "10",        // 语音时长，单位：秒
+  "fileSize": "2048"       // 可选，单位：字节
+}
+```
+
+**ActionCard 单按钮 (`sampleActionCard`)**
+
+```json
+{
+  "title": "卡片标题",
+  "markdown": "#### 内容标题\n正文",
+  "singleTitle": "查看详情",
+  "singleUrl": "https://example.com"
+}
+```
+
+**ActionCard 多按钮 (`sampleMultiActionCard`)**
+
+```json
+{
+  "title": "卡片标题",
+  "markdown": "#### 内容标题\n正文",
+  "btnOrientation": "0",   // "0"=竖排，"1"=横排
+  "btns": [
+    {"title": "同意", "url": "https://example.com/approve"},
+    {"title": "拒绝", "url": "https://example.com/reject"}
+  ]
+}
+```
+
+#### 响应参数
+
+| 参数名 | 类型 | 说明 |
+|--------|------|------|
+| `processQueryKey` | String | 消息发送任务的查询 Key，可用于查询发送结果 |
+
+**成功响应示例：**
+
+```json
+{
+  "processQueryKey": "msgTaskId_xxx"
+}
+```
+
+**错误响应示例：**
+
+```json
+{
+  "code": "InvalidParameter.RobotCode",
+  "message": "robotCode is invalid",
+  "requestid": "xxxx-xxxx-xxxx"
+}
+```
+
+#### 请求示例（cURL）
+
+```bash
+curl -X POST 'https://api.dingtalk.com/v1.0/robot/oToMessages/batchSend' \
+  -H 'x-acs-dingtalk-access-token: YOUR_ACCESS_TOKEN' \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "robotCode": "dingxxxxxxxx",
+    "userIds": ["user001", "user002"],
+    "msgKey": "sampleText",
+    "msgParam": "{\"content\": \"Hello, 这是一条测试消息！\"}"
+  }'
+```
+
+#### 请求示例（Python）
+
+```python
+import requests
+import json
+
+url = "https://api.dingtalk.com/v1.0/robot/oToMessages/batchSend"
+headers = {
+    "x-acs-dingtalk-access-token": "YOUR_ACCESS_TOKEN",
+    "Content-Type": "application/json"
+}
+payload = {
+    "robotCode": "dingxxxxxxxx",
+    "userIds": ["user001", "user002"],
+    "msgKey": "sampleText",
+    "msgParam": json.dumps({"content": "Hello, 这是一条测试消息！"})
+}
+
+response = requests.post(url, headers=headers, json=payload)
+print(response.json())
+```
+
+---
+
+### 2. 发送群聊消息
+
+> 官方文档：https://open.dingtalk.com/document/orgapp/the-robot-sends-a-group-message
+
+#### 接口概述
+
+调用本接口向指定群聊发送机器人消息。机器人需要先加入群聊才能发送消息。
+
+#### 请求方式
+
+- **HTTP 方法**：`POST`
+- **URL**：`https://api.dingtalk.com/v1.0/robot/groupMessages/send`
+
+#### 请求头
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `x-acs-dingtalk-access-token` | String | 是 | access_token |
+| `Content-Type` | String | 是 | 固定值：`application/json` |
+
+#### 请求体参数（Body）
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `robotCode` | String | 是 | 机器人编码 |
+| `openConversationId` | String | 是 | 群聊会话 ID |
+| `msgKey` | String | 是 | 消息模板 Key |
+| `msgParam` | String | 是 | 消息模板参数，JSON 字符串 |
+
+#### 响应参数
+
+| 参数名 | 类型 | 说明 |
+|--------|------|------|
+| `processQueryKey` | String | 消息发送任务的查询 Key |
+
+---
+
+### 3. 获取 Access Token
+
+> 官方文档：https://open.dingtalk.com/document/orgapp/obtain-the-access_token-of-an-internal-app
+
+#### 请求方式
+
+- **HTTP 方法**：`GET`
+- **URL**：`https://oapi.dingtalk.com/gettoken`
+
+#### 查询参数（Query）
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `appkey` | String | 是 | 应用的 AppKey |
+| `appsecret` | String | 是 | 应用的 AppSecret |
+
+#### 响应参数
+
+| 参数名 | 类型 | 说明 |
+|--------|------|------|
+| `errcode` | Number | 错误码，0 表示成功 |
+| `errmsg` | String | 错误信息 |
+| `access_token` | String | 访问凭证 |
+| `expires_in` | Number | 有效期（秒），通常为 7200 |
+
+#### 响应示例
+
+```json
+{
+  "errcode": 0,
+  "errmsg": "ok",
+  "access_token": "xxxxxx",
+  "expires_in": 7200
+}
+```
+
+---
+
+### 4. 上传媒体文件
+
+> 用于发送图片、语音、文件类消息前的文件上传。
+
+#### 请求方式
+
+- **HTTP 方法**：`POST`
+- **URL**：`https://oapi.dingtalk.com/media/upload`
+
+#### 查询参数（Query）
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `access_token` | String | 是 | 访问凭证 |
+
+#### 请求体参数（multipart/form-data）
+
+| 参数名 | 类型 | 必填 | 说明 |
+|--------|------|------|------|
+| `type` | String | 是 | 文件类型：`image`/`voice`/`file` |
+| `media` | File | 是 | 上传的文件 |
+
+#### 响应参数
+
+| 参数名 | 类型 | 说明 |
+|--------|------|------|
+| `errcode` | Number | 错误码 |
+| `errmsg` | String | 错误信息 |
+| `media_id` | String | 媒体文件 ID，后续发送消息时使用 |
+| `type` | String | 文件类型 |
+| `created_at` | Number | 创建时间戳（毫秒） |
+
+---
+
+### 二、Webhook 自定义机器人 API
+
+> 官方文档：https://open.dingtalk.com/document/orgapp/custom-bot-to-send-group-chat-messages
+
+#### Webhook 概述
+
+自定义机器人是一种可以直接添加到钉钉群聊的机器人，通过 Webhook URL 推送消息到群聊。与企业内部机器人不同，不需要创建应用、不需要 OAuth Token 流程，只需 POST JSON 到 Webhook URL 即可。
+
+**适用场景：**
+- 监控告警通知
+- CI/CD 构建通知
+- 定时数据报告推送
+- 简单的群聊消息推送
+
+**Webhook URL 格式：**
+
+```
+https://oapi.dingtalk.com/robot/send?access_token=XXXXXX
+```
+
+#### 请求方式
+
+- **HTTP 方法**：`POST`
+- **Content-Type**：`application/json; charset=utf-8`
+- **频率限制**：每个机器人每分钟最多发送 **20** 条消息
+
+#### 安全设置
+
+创建自定义机器人时，必须至少选择以下三种安全方式之一：
+
+**方式一：自定义关键词**
+
+- 最多设置 10 个关键词
+- 消息内容必须包含至少一个关键词，否则被拒绝
+
+**方式二：IP 地址白名单**
+
+- 配置允许的 IP 地址或 CIDR 段
+- 仅允许白名单内的 IP 发送消息
+
+**方式三：加签（推荐）**
+
+启用加签后，每次请求需要在 URL 中附加 `timestamp` 和 `sign` 参数。
+
+**签名算法：**
+
+```
+timestamp = 当前毫秒时间戳
+string_to_sign = timestamp + "\n" + secret
+sign = URL_Encode(Base64(HMAC-SHA256(secret, string_to_sign)))
+```
+
+**签名后的 URL：**
+
+```
+https://oapi.dingtalk.com/robot/send?access_token=XXXXXX&timestamp=1609459200000&sign=YYYY
+```
+
+> 时间戳与服务器时间差不能超过 1 小时，否则请求被拒绝。
+> 脚本已内置加签逻辑，只需在配置文件中设置 `webhook_secret` 即可自动签名。
+
+**Python 签名示例：**
+
+```python
+import time, hmac, hashlib, base64, urllib.parse
+
+def generate_sign(secret: str):
+    timestamp = str(round(time.time() * 1000))
+    string_to_sign = f'{timestamp}\n{secret}'
+    hmac_code = hmac.new(
+        secret.encode('utf-8'),
+        string_to_sign.encode('utf-8'),
+        digestmod=hashlib.sha256
+    ).digest()
+    sign = urllib.parse.quote_plus(base64.b64encode(hmac_code))
+    return timestamp, sign
+```
+
+#### Webhook 消息类型
+
+##### 文本消息 (text)
+
+```json
+{
+    "msgtype": "text",
+    "text": {
+        "content": "消息内容"
+    },
+    "at": {
+        "atMobiles": ["13800138000"],
+        "atUserIds": ["user123"],
+        "isAtAll": false
+    }
+}
+```
+
+##### Markdown 消息 (markdown)
+
+```json
+{
+    "msgtype": "markdown",
+    "markdown": {
+        "title": "标题（通知栏显示）",
+        "text": "#### 标题\n> 引用内容\n正文"
+    },
+    "at": {
+        "atMobiles": ["13800138000"],
+        "isAtAll": false
+    }
+}
+```
+
+##### 链接消息 (link)
+
+> 不支持 @功能
+
+```json
+{
+    "msgtype": "link",
+    "link": {
+        "title": "链接标题",
+        "text": "链接描述",
+        "messageUrl": "https://example.com",
+        "picUrl": "https://example.com/image.png"
+    }
+}
+```
+
+##### ActionCard 单按钮 (actionCard)
+
+```json
+{
+    "msgtype": "actionCard",
+    "actionCard": {
+        "title": "卡片标题",
+        "text": "#### 内容\n正文（支持 Markdown）",
+        "btnOrientation": "0",
+        "singleTitle": "阅读全文",
+        "singleURL": "https://example.com/"
+    }
+}
+```
+
+##### ActionCard 多按钮 (actionCard)
+
+```json
+{
+    "msgtype": "actionCard",
+    "actionCard": {
+        "title": "卡片标题",
+        "text": "#### 内容\n正文",
+        "btnOrientation": "0",
+        "btns": [
+            {"title": "同意", "actionURL": "https://example.com/approve"},
+            {"title": "拒绝", "actionURL": "https://example.com/reject"}
+        ]
+    }
+}
+```
+
+> 注意：多按钮时使用 `btns` + `actionURL`，不要同时设置 `singleTitle`/`singleURL`
+
+##### FeedCard 消息 (feedCard)
+
+> 不支持 @功能
+
+```json
+{
+    "msgtype": "feedCard",
+    "feedCard": {
+        "links": [
+            {
+                "title": "链接标题1",
+                "messageURL": "https://example.com/1",
+                "picURL": "https://example.com/pic1.png"
+            },
+            {
+                "title": "链接标题2",
+                "messageURL": "https://example.com/2",
+                "picURL": "https://example.com/pic2.png"
+            }
+        ]
+    }
+}
+```
+
+#### Webhook 响应
+
+**成功：**
+
+```json
+{
+    "errcode": 0,
+    "errmsg": "ok"
+}
+```
+
+**错误示例：**
+
+```json
+{
+    "errcode": 310000,
+    "errmsg": "keywords not in content"
+}
+```
+
+#### 请求示例（cURL）
+
+```bash
+curl -X POST 'https://oapi.dingtalk.com/robot/send?access_token=XXXXXX' \
+  -H 'Content-Type: application/json; charset=utf-8' \
+  -d '{
+    "msgtype": "text",
+    "text": {
+        "content": "Hello, 这是一条 Webhook 测试消息！"
+    }
+  }'
+```
+
+#### Webhook 与企业内部机器人对比
+
+| 维度 | Webhook 自定义机器人 | 企业内部机器人 |
+|------|---------------------|---------------|
+| **接入方式** | 群聊添加自定义机器人 | 钉钉开放平台创建应用 |
+| **API 域名** | `oapi.dingtalk.com/robot/send` | `api.dingtalk.com/v1.0/robot/` |
+| **认证方式** | URL 中的 access_token + 可选签名 | Header 中的 OAuth access_token |
+| **配置项** | webhook_url, webhook_secret | app_key, app_secret, robot_code |
+| **消息范围** | 仅所在群聊 | 任意用户/群聊 |
+| **消息格式** | `msgtype` + 类型对象 | `msgKey` + `msgParam` (JSON 字符串) |
+| **频率限制** | 20 条/分钟 | 20 次/秒 |
+| **Token 管理** | 无需刷新（URL 固定） | Token 每 2 小时过期 |
+
+---
+
+## CLI 使用指南
+
+### 安装依赖
+
+```bash
+pip install requests
+```
+
+### 命令行格式
+
+```bash
+# 企业内部机器人
+python scripts/dingtalk.py <消息类型> [选项参数] "<消息内容>" --users <用户ID>
+
+# Webhook 自定义机器人（命令以 webhook- 前缀开头）
+python scripts/dingtalk.py webhook-<消息类型> [选项参数] "<消息内容>"
+```
+
+### 企业内部机器人命令
+
+#### 发送文本消息
+
+```bash
+# 单聊
+python scripts/dingtalk.py text "Hello, 这是一条测试消息！" --users user001,user002
+
+# @指定用户
+python scripts/dingtalk.py text "Hello, @user001" --users user001,user002 --at-users user001
+
+# 群聊
+python scripts/dingtalk.py text "大家好！" \
+    --mode group \
+    --conversation-id chatxxxxxxxxxxxxxxxx \
+    --at-mobiles 13800138000,13900139000
+```
+
+#### 发送 Markdown 消息
+
+```bash
+python scripts/dingtalk.py markdown \
+    --title "天气提醒" \
+    --users user001 \
+    "#### 杭州天气\n> 9度，西北风1级"
+```
+
+#### 发送链接消息
+
+```bash
+python scripts/dingtalk.py link \
+    --title "时代在进步" \
+    --url "https://www.dingtalk.com" \
+    --pic-url "https://example.com/image.png" \
+    --users user001 \
+    "点击查看详情"
+```
+
+#### 发送 ActionCard 消息
+
+```bash
+# 单按钮
+python scripts/dingtalk.py action-card \
+    --title "审批通知" \
+    --single-title "查看详情" \
+    --url "https://www.dingtalk.com" \
+    --users user001 \
+    "#### 请假申请\n请审批"
+
+# 多按钮
+python scripts/dingtalk.py action-card \
+    --title "审批通知" \
+    --buttons "同意,https://approve.com/yes;拒绝,https://approve.com/no" \
+    --btn-orientation 0 \
+    --users user001 \
+    "#### 请假申请"
+```
+
+#### 发送文件
+
+```bash
+python scripts/dingtalk.py file \
+    --users user001,user002 \
+    --file /path/to/report.pdf \
+    --file-name "月度报告.pdf"
+```
+
+### Webhook 自定义机器人命令
+
+> 所有 Webhook 命令以 `webhook-` 前缀开头。
+> Webhook URL 可通过 `--webhook-url` 参数传入，或在配置文件中设置 `webhook_url`。
+
+#### Webhook 发送文本消息
+
+```bash
+# 使用配置文件中的 webhook_url
+python scripts/dingtalk.py webhook-text "Hello, 这是一条 Webhook 消息！"
+
+# 通过参数指定 webhook URL
+python scripts/dingtalk.py webhook-text \
+    --webhook-url "https://oapi.dingtalk.com/robot/send?access_token=xxx" \
+    "Hello, 测试消息"
+
+# @指定手机号
+python scripts/dingtalk.py webhook-text \
+    --at-mobiles 13800138000,13900139000 \
+    "通知内容 @13800138000"
+
+# @所有人
+python scripts/dingtalk.py webhook-text --at-all "全员通知"
+```
+
+#### Webhook 发送 Markdown 消息
+
+```bash
+python scripts/dingtalk.py webhook-markdown \
+    --title "天气提醒" \
+    "#### 杭州天气\n> 9度，西北风1级"
+```
+
+#### Webhook 发送链接消息
+
+```bash
+python scripts/dingtalk.py webhook-link \
+    --title "时代在进步" \
+    --url "https://www.dingtalk.com" \
+    --pic-url "https://example.com/image.png" \
+    "点击查看详情"
+```
+
+#### Webhook 发送 ActionCard 消息
+
+```bash
+# 单按钮
+python scripts/dingtalk.py webhook-action-card \
+    --title "审批通知" \
+    --single-title "查看详情" \
+    --url "https://www.dingtalk.com" \
+    "#### 请假申请\n请审批"
+
+# 多按钮
+python scripts/dingtalk.py webhook-action-card \
+    --title "审批通知" \
+    --buttons "同意,https://approve.com/yes;拒绝,https://approve.com/no" \
+    --btn-orientation 0 \
+    "#### 请假申请"
+```
+
+#### Webhook 发送 FeedCard 消息
+
+```bash
+python scripts/dingtalk.py webhook-feed-card \
+    --links "新闻标题1,https://news1.com,https://img1.com/pic.png;新闻标题2,https://news2.com,https://img2.com/pic.png"
+```
+
+#### Webhook 使用加签
+
+```bash
+# 通过命令行参数
+python scripts/dingtalk.py webhook-text \
+    --webhook-url "https://oapi.dingtalk.com/robot/send?access_token=xxx" \
+    --webhook-secret "SECxxxxxxxxxxxxxxxxxxxxxxxxxx" \
+    "带签名的消息"
+
+# 或在配置文件中设置（推荐）
+python scripts/dingtalk.py config --set webhook_secret=SECxxxxxxxxxxxxxxxxxxxxxxxxxx
+```
+
+### 查看帮助
+
+```bash
+python scripts/dingtalk.py --help
+python scripts/dingtalk.py text --help
+python scripts/dingtalk.py webhook-text --help
+python scripts/dingtalk.py webhook-markdown --help
+```
+
+---
+
+## 配置说明
+
+### 配置文件
+
+配置统一存储在系统配置目录，所有 AI agent 共享，无需重复配置：
+
+| 平台 | 配置文件 | 状态文件 |
+|------|---------|---------|
+| macOS / Linux | `~/.config/dingtalk/config.json` | `~/.config/dingtalk/state.json` |
+| Windows | `%APPDATA%\dingtalk\config.json` | `%APPDATA%\dingtalk\state.json` |
+
+```json
+{
+    "default_robot": "技术告警群",
+    "robots": [
+        {
+            "name": "技术告警群",
+            "type": "webhook",
+            "description": "发送技术告警到后端技术群",
+            "webhook_token": "YOUR_ACCESS_TOKEN",
+            "webhook_secret": ""
+        },
+        {
+            "name": "内部通知机器人",
+            "type": "app",
+            "description": "企业内部机器人，支持单聊和群聊",
+            "app_key": "YOUR_APP_KEY",
+            "app_secret": "YOUR_APP_SECRET",
+            "robot_code": "YOUR_ROBOT_CODE",
+            "agent_id": "YOUR_AGENT_ID"
+        }
+    ]
+}
+```
+
+### 参数获取说明
+
+#### 企业内部机器人
+
+| 参数 | 说明 | 获取方式 | 必填 |
+|------|------|----------|------|
+| `app_key` | 应用 AppKey | 钉钉开放平台 > 应用详情 > 凭证与基础信息 | 是 |
+| `app_secret` | 应用 AppSecret | 钉钉开放平台 > 应用详情 > 凭证与基础信息 | 是 |
+| `robot_code` | 机器人编号 | 钉钉开放平台 > 应用详情 > 机器人与消息推送 | 是 |
+| `agent_id` | 应用 AgentID | 钉钉开放平台 > 应用详情 > 凭证与基础信息 | 否 |
+
+#### Webhook 自定义机器人
+
+| 参数 | 说明 | 获取方式 | 必填 |
+|------|------|----------|------|
+| `webhook_url` | Webhook 地址 | 钉钉群 > 群设置 > 智能群助手 > 添加自定义机器人 > 复制 Webhook | 是 |
+| `webhook_secret` | 加签密钥 | 创建机器人时选择"加签"安全方式，复制 SEC 开头的密钥 | 否* |
+
+> *如果安全设置选择了"加签"方式，则 webhook_secret 必填。
+> 两种机器人的配置可以同时存在于同一个配置文件中，互不影响。
+
+### 配置管理命令
+
+```bash
+# 初始化配置文件
+python scripts/dingtalk.py config --init
+
+# 查看当前配置
+python scripts/dingtalk.py config --show
+
+# 设置单个配置项
+python scripts/dingtalk.py config --set app_key=dingxxxxxxxxxxxx
+
+# 添加机器人
+python scripts/dingtalk.py robot-add --name "技术告警群" --type webhook --webhook-token "token_xxx"
+python scripts/dingtalk.py robot-add --name "内部通知" --type app --app-key dingxxx --app-secret xxx --robot-code robot-xxx
+
+# 强制覆盖已存在的配置
+python scripts/dingtalk.py config --init --force
+```
+
+### 命令行参数覆盖
+
+配置文件中的值可以通过命令行参数临时覆盖：
+
+```bash
+python scripts/dingtalk.py text "测试消息" \
+    --users user001 \
+    --app-key dingxxxxxxxxxxxx \
+    --app-secret your-secret-key \
+    --robot-code robot-xxxxxx
+```
+
+---
+
+## 消息类型详解
+
+| 类型 | CLI 指令 | 模式 | 说明 |
+|------|----------|------|------|
+| 文本消息 | `text` | 单聊/群聊 | 纯文本，支持 @用户 |
+| Markdown | `markdown` | 单聊/群聊 | 有限 Markdown 语法的富文本 |
+| 链接消息 | `link` | 单聊/群聊 | 带缩略图的链接卡片 |
+| ActionCard | `action-card` | 单聊/群聊 | 带按钮的交互卡片（单/多按钮） |
+| 文件消息 | `file` | 仅单聊 | 发送文件，自动上传 |
+| 图片消息 | `image` | 仅单聊 | 发送图片 |
+| 语音消息 | `voice` | 仅单聊 | 发送语音 |
+
+
+---
+
+## 钉钉 Markdown 语法限制
+
+钉钉的 Markdown 渲染器只支持标准 Markdown 的有限子集。
+
+### 支持的语法
+
+| 语法 | 写法 | 说明 |
+|------|------|------|
+| 标题 | `# 一级标题` ~ `###### 六级标题` | 正常支持 |
+| 加粗 | `**粗体文字**` | 正常支持 |
+| 链接 | `[链接文字](url)` | 正常支持 |
+| 图片 | `![alt](图片url)` | 正常支持 |
+| 无序列表 | `- 列表项` | 正常支持 |
+| 有序列表 | `1. 列表项` | 正常支持 |
+| 引用 | `> 引用文字` | 正常支持 |
+
+### 不支持的语法（禁止使用）
+
+以下语法在钉钉中**不会被渲染**，甚至可能导致显示异常：
+
+- 水平分隔线：`---`、`***`、`___`
+- 表格：`| col1 | col2 |`
+- 代码块：`` ``` ``
+- 行内代码：`` `code` ``
+- 删除线：`~~text~~`
+- 任务列表：`- [ ] item`
+- 斜体：`*italic*`（不稳定）
+- 嵌套列表：多级缩进列表（显示不稳定）
+
+### 编写建议
+
+1. 用 `#` ~ `####` 标题组织结构
+2. 用 `- item` 无序列表展示条目
+3. 用 `**重点**` 加粗强调关键信息
+4. 用 `> 备注` 添加补充说明
+5. 不要用分隔线，用标题或空行替代
+6. 不要用表格，用列表格式替代
+7. 在 CLI 中使用 `\n` 表示换行
+
+---
+
+## 错误码参考
+
+### 企业内部机器人
+
+| 错误码 | 说明 |
+|--------|------|
+| 0 | 成功 |
+| -1 | 系统繁忙，请稍后重试 |
+| 40001 | access_token 不存在或已过期 |
+| 40002 | access_token 不合法 |
+| 40004 | 无效的机器人（robotCode 错误） |
+| 40007 | 无效的 openConversationId |
+| 40008 | 无效的消息内容（msgParam 格式错误） |
+| 40009 | 消息发送失败，用户不在该机器人可见范围 |
+| 40010 | 消息发送失败，用户未与机器人建立会话 |
+| 40014 | 无效的 userId |
+| 40037 | 发送消息过于频繁，已触发限流 |
+| 40056 | 无效的 agentId |
+
+### Webhook 自定义机器人
+
+| 错误码 | 说明 |
+|--------|------|
+| 0 | 成功 |
+| 300001 | 无效的 token / 机器人不存在 |
+| 310000 | 安全校验失败（关键词不匹配 / IP 不在白名单 / 签名错误 / 时间戳过期） |
+| 302503 | 频率限制，每分钟超过 20 条 |
+| 400013 | JSON 格式无效 |
+
+---
+
+## 注意事项与限制
+
+### 企业内部机器人
+
+#### 频率限制
+
+- 每个应用每秒钟最多调用 **20 次** 消息发送接口
+- 超出限制会返回错误码 `40037`
+
+#### 用户限制
+
+- 批量发送单聊消息时，`userIds` 每次最多 **100 个**
+- 超过 100 个需分批发送
+
+#### 会话前置条件
+
+- 发送单聊消息前，用户需要先与机器人**建立会话**（用户主动给机器人发送过消息）
+- 未建立会话的用户会收到错误码 `40010`
+
+#### 文件上传
+
+- 发送文件/图片/语音消息前，需要先通过 `/media/upload` 接口上传文件获取 `media_id`
+- CLI 工具的 `file` 命令已自动集成上传流程
+
+#### Token 管理
+
+- `access_token` 有效期为 **7200 秒**（2 小时）
+- 工具内置缓存机制，过期前 5 分钟自动刷新
+- 不建议频繁调用 gettoken 接口
+
+### Webhook 自定义机器人
+
+#### 频率限制
+
+- 每个机器人每分钟最多发送 **20 条** 消息
+- 超出限制返回错误码 `302503`
+
+#### 安全设置
+
+- 创建机器人时必须至少选择一种安全方式：自定义关键词、IP 白名单、加签
+- 使用加签时，时间戳与服务器时间差不能超过 **1 小时**
+- 脚本内置加签逻辑，配置 `webhook_secret` 后自动签名
+
+#### 消息范围
+
+- Webhook 机器人只能在添加到的群聊中发送消息，**不支持单聊**
+- `link` 和 `feedCard` 消息类型**不支持** @功能
+
+#### Token 管理
+
+- Webhook URL 中的 `access_token` 是固定的，无需刷新
+- 只要机器人未被删除，URL 持续有效
+
+### 通用
+
+#### 消息内容
+
+- Markdown 消息仅支持有限语法子集，见 [Markdown 语法限制](#钉钉-markdown-语法限制)
+- 消息内容不宜过长，过长建议拆分或使用文件发送
+- 企业内部机器人的 `msgParam` 需要序列化为 JSON 字符串后传入（脚本已自动处理）
+
+#### API 域名
+
+- 企业内部机器人（消息发送）：`https://api.dingtalk.com`
+- 企业内部机器人（gettoken、文件上传）：`https://oapi.dingtalk.com`
+- Webhook 自定义机器人：`https://oapi.dingtalk.com/robot/send`
--- a/skills/dingtalk-message/SKILL.md
+++ b/skills/dingtalk-message/SKILL.md
@ -0,0 +1,268 @@
+---
+name: dingtalk-message
+version: 0.3.0
+description: 钉钉消息发送技能。支持企业内部机器人（批量单聊/群聊）和 Webhook 自定义机器人两种接入方式，支持多机器人管理，支持文本、Markdown、链接、ActionCard、FeedCard等多种消息类型。
+---
+
+# 钉钉消息发送技能
+
+## 概述
+
+支持两种接入方式：
+- **Webhook 自定义机器人**：通过 access_token 向群聊发送消息，接入简单
+- **企业内部机器人**：通过 app_key/app_secret 发送单聊、群聊消息，功能更全
+
+支持多机器人管理，只有一个时自动使用，多个时按优先级自动选择。
+
+## 环境要求
+
+- Python 3.7+
+- `pip install requests`
+
+## 首次配置
+
+### 引导流程
+
+首次使用此技能时，必须按以下流程引导用户完成配置：
+
+1. **询问用户的机器人类型和凭证信息**：
+   - Webhook 机器人：需要 `access_token`（和可选的加签密钥 `secret`）
+   - 企业内部机器人：需要 `app_key`、`app_secret`、`robot_code`（和可选的 `agent_id`）
+
+2. **执行配置命令**：`python scripts/dingtalk.py robot-add --name "机器人名" --type webhook ...`
+
+3. **验证配置**：`python scripts/dingtalk.py config --show`
+
+### 配置文件路径
+
+配置统一存储在系统配置目录，所有 AI agent 共享，无需重复配置：
+
+| 平台 | 配置文件 | 状态文件 |
+|------|---------|---------|
+| macOS / Linux | `~/.config/dingtalk/config.json` | `~/.config/dingtalk/state.json` |
+| Windows | `%APPDATA%\dingtalk\config.json` | `%APPDATA%\dingtalk\state.json` |
+
+### 手动编辑配置文件
+
+也可直接编辑配置文件，每个机器人用 `name` 标识（建议用群名、用途等有意义的名称），`description` 描述用途，方便智能匹配：
+
+```json
+{
+    "default_robot": "技术告警群",
+    "robots": [
+        {
+            "name": "技术告警群",
+            "type": "webhook",
+            "description": "发送技术告警到后端技术群",
+            "webhook_token": "你的access_token",
+            "webhook_secret": ""
+        },
+        {
+            "name": "内部通知机器人",
+            "type": "app",
+            "description": "企业内部机器人，支持单聊和群聊",
+            "app_key": "你的AppKey",
+            "app_secret": "你的AppSecret",
+            "robot_code": "你的机器人编号",
+            "agent_id": ""
+        }
+    ]
+}
+```
+
+> `webhook_token` 填 access_token 即可，脚本自动拼接完整 URL。
+
+### 通过命令行添加
+
+```bash
+# 添加 Webhook 机器人
+python scripts/dingtalk.py robot-add --name "技术告警群" --type webhook --webhook-token "access_token_xxx" --desc "发送告警到后端技术群"
+
+# 添加企业内部机器人
+python scripts/dingtalk.py robot-add --name "内部通知" --type app --app-key dingxxx --app-secret xxx --robot-code robot-xxx --desc "支持单聊群聊"
+```
+
+### 验证配置
+
+```bash
+python scripts/dingtalk.py config --show
+```
+
+## 机器人管理
+
+### 选择逻辑
+
+- 只配置一个机器人时，自动使用
+- 多个机器人时：`--robot` 指定 > `default_robot` > 最近使用过的 > 第一个可用的
+- 用户未明确指定时，可根据机器人的 `description` 和最近消息记录智能匹配，或询问用户
+
+### 管理命令
+
+```bash
+# 添加（--desc 描述用途，便于记忆和智能选择）
+python scripts/dingtalk.py robot-add --name "技术告警群" --type webhook --webhook-token "access_token_xxx" --desc "发送告警到后端技术群"
+python scripts/dingtalk.py robot-add --name "内部通知" --type app --app-key dingxxx --app-secret xxx --robot-code robot-xxx --desc "企业内部机器人，支持单聊群聊"
+
+# 查看所有机器人（含描述、使用次数、最近消息）
+python scripts/dingtalk.py robot-list
+
+# 更新描述 / 重命名
+python scripts/dingtalk.py robot-update --name "技术告警群" --desc "后端+SRE告警群"
+python scripts/dingtalk.py robot-update --name "alert-bot" --rename "技术告警群"
+
+# 指定机器人发送
+python scripts/dingtalk.py webhook-text --robot "技术告警群" "告警消息"
+
+# 设置默认 / 启用禁用 / 删除
+python scripts/dingtalk.py robot-default --name "产品日报群"
+python scripts/dingtalk.py robot-enable --name "技术告警群" --disable
+python scripts/dingtalk.py robot-remove --name "技术告警群"
+```
+
+### 使用记录
+
+每次发送消息会自动记录摘要到状态文件，包括：
+- 使用次数、最近使用时间、最近状态
+- 最近10条消息摘要（消息类型 + 内容前60字）
+
+通过 `robot-list` 可查看，也用于智能选择机器人。
+
+## Webhook 消息
+
+> Webhook 命令以 `webhook-` 前缀开头。
+> Webhook URL 获取：钉钉群 > 群设置 > 智能群助手 > 添加自定义机器人 > 复制 Webhook 地址
+
+### 文本消息
+
+```bash
+# 已配置时直接发送
+python scripts/dingtalk.py webhook-text "消息内容"
+
+# 临时指定 token（无需配置）
+python scripts/dingtalk.py webhook-text --webhook-token "access_token_xxx" "消息内容"
+
+# @用户 / @所有人
+python scripts/dingtalk.py webhook-text --at-mobiles 13800138000 "消息 @13800138000"
+python scripts/dingtalk.py webhook-text --at-all "全员通知"
+```
+
+### Markdown 消息
+
+```bash
+python scripts/dingtalk.py webhook-markdown \
+    --title "天气提醒" \
+    "#### 杭州天气\n> 9度，西北风1级"
+```
+
+### 链接消息
+
+```bash
+python scripts/dingtalk.py webhook-link \
+    --title "时代在进步" \
+    --url "https://www.dingtalk.com" \
+    --pic-url "https://example.com/image.png" \
+    "点击查看详情"
+```
+
+### ActionCard 消息
+
+```bash
+# 单按钮
+python scripts/dingtalk.py webhook-action-card \
+    --title "审批通知" \
+    --single-title "查看详情" \
+    --url "https://www.dingtalk.com" \
+    "#### 请假申请\n请审批"
+
+# 多按钮
+python scripts/dingtalk.py webhook-action-card \
+    --title "审批通知" \
+    --buttons "同意,https://approve.com/yes;拒绝,https://approve.com/no" \
+    --btn-orientation 0 \
+    "#### 请假申请"
+```
+
+### FeedCard 消息
+
+```bash
+python scripts/dingtalk.py webhook-feed-card \
+    --links "新闻1,https://news1.com,https://img1.com/pic.png;新闻2,https://news2.com,https://img2.com/pic.png"
+```
+
+### 加签安全
+
+```bash
+# 配置加签密钥（一次设置）
+python scripts/dingtalk.py config --set webhook_secret=SECxxxxxxxxxxxxxxxxxxxxxxxxxx
+
+# 或通过参数临时指定
+python scripts/dingtalk.py webhook-text \
+    --webhook-token "access_token_xxx" \
+    --webhook-secret "SECxxxxxxxxxxxxxxxxxxxxxxxxxx" \
+    "带加签的消息"
+```
+
+## 企业内部机器人消息
+
+### 单聊消息
+
+```bash
+# 文本
+python scripts/dingtalk.py text "Hello!" --users user001,user002
+
+# Markdown
+python scripts/dingtalk.py markdown --title "天气提醒" --users user001 \
+    "#### 杭州天气\n> 9度，西北风1级"
+
+# 链接
+python scripts/dingtalk.py link --title "时代在进步" --url "https://www.dingtalk.com" \
+    --users user001 "点击查看详情"
+
+# ActionCard
+python scripts/dingtalk.py action-card --title "审批通知" \
+    --single-title "查看详情" --url "https://www.dingtalk.com" \
+    --users user001 "#### 请假申请\n请审批"
+
+# 文件
+python scripts/dingtalk.py file --users user001 --file /path/to/report.pdf --file-name "月度报告.pdf"
+```
+
+### 群聊消息
+
+```bash
+python scripts/dingtalk.py text "大家好！" \
+    --mode group \
+    --conversation-id chatxxxxxxxxxxxxxxxx \
+    --at-mobiles 13800138000,13900139000
+```
+
+## CLI 语法规则
+
+1. `content` 是位置参数（无 `--` 前缀），放在可选参数之后
+2. 换行使用 `\n`，脚本自动处理转换
+3. `content` 必须用双引号包裹
+4. 可通过 `--app-key`、`--robot-code` 等参数临时覆盖配置文件
+
+## 钉钉 Markdown 语法限制
+
+钉钉 Markdown 只支持有限子集，构造内容时必须遵守：
+
+**支持：** 标题(`#`)、加粗(`**`)、链接(`[](url)`)、图片(`![](url)`)、无序列表(`-`)、有序列表(`1.`)、引用(`>`)
+
+**不支持（禁止使用）：** 分隔线(`---`)、表格、代码块、行内代码、删除线、任务列表、斜体、嵌套列表
+
+> 脚本会自动移除分隔线 `---`，但其他不支持的语法需要手动避免。
+
+## 常见错误码
+
+**企业内部机器人：** 40001(token过期) / 40004(无效机器人) / 40009(用户不在可见范围) / 40010(未建立会话) / 40037(发送过频)
+
+**Webhook：** 300001(无效token) / 310000(签名校验失败) / 302503(频率限制，每分钟20条)
+
+## 注意事项
+
+- **频率限制**：企业内部机器人每秒20次，Webhook每分钟20条
+- **Webhook 仅群聊**：不支持单聊，link/feedCard 不支持 @功能
+- **单聊前置条件**：用户需先主动给机器人发过消息
+- **批量限制**：单次最多发送100个用户
+- **消息长度**：过长内容建议拆分或使用文件发送
--- a/skills/dingtalk-message/package.json
+++ b/skills/dingtalk-message/package.json
@ -0,0 +1,11 @@
+{
+  "name": "dingtalk-message",
+  "version": "0.3.0",
+  "description": "钉钉消息发送技能。支持企业内部机器人（批量单聊/群聊）和 Webhook 自定义机器人两种接入方式，支持多机器人管理，支持文本、Markdown、链接、ActionCard、FeedCard等多种消息类型。",
+  "publishConfig": {
+    "registry": "https://contextlab.alibaba-inc.com/skill"
+  },
+  "aoneKit": {
+    "generated": true
+  }
+}
--- a/skills/dingtalk-message/scripts/dingtalk.py
+++ b/skills/dingtalk-message/scripts/dingtalk.py
--- a/skills/serper-search/SKILL.md
+++ b/skills/serper-search/SKILL.md
@ -0,0 +1,45 @@
+---
+name: serper-search
+description: Search the public web through Serper instead of the built-in WebSearch tool. Use when an agent needs current information, general web sources, or URLs not available from local files and project APIs. SERPER_API_KEY is available in the environment.
+---
+
+# Serper Search
+
+Use Serper whenever external web search is needed. Prefer local files and project APIs when they already contain the required information.
+
+## Rules
+
+- Do not use the built-in WebSearch tool.
+- Do not expose the API key in answers, logs, committed fixtures, or browser/client code.
+- Prefer authoritative and primary sources in the returned results.
+- Refine the query when freshness, date, company, product, or source type matters.
+
+## API
+
+- Endpoint: `POST https://google.serper.dev/search`
+- Auth header: `X-API-KEY: $SERPER_API_KEY`
+- Content type: `application/json`
+- Required body: `{"q":"query text"}`
+
+## Curl
+
+```bash
+curl -s --location "https://google.serper.dev/search" \
+  --header "X-API-KEY: $SERPER_API_KEY" \
+  --header "Content-Type: application/json" \
+  --data '{"q":"apple inc"}'
+```
+
+## Response Use
+
+- Prefer `organic` results for normal search answers.
+- Use `searchParameters` to confirm the query actually sent.
+- Treat `credits` as usage accounting.
+- Verify important claims against the linked source rather than treating snippets as complete evidence.
+- Summarize results with source titles and links; do not copy large passages from snippets or pages.
+
+## Failure Handling
+
+1. If DNS/network fails, report it as an environment/network issue and retry only when external network is available.
+2. If authentication fails, report that `SERPER_API_KEY` is missing or invalid without printing its value.
+3. If results are irrelevant, refine the query with names, dates, locations, or source domains.
--- a/tests/unit/test_base_agent_wrapper.py
+++ b/tests/unit/test_base_agent_wrapper.py
@ -1,11 +1,14 @@
 """Tests for shared agent wrapper behavior."""

+import sys
 from unittest.mock import MagicMock

 import pytest

 from reme.components.agent_wrapper import AsAgentWrapper, BaseAgentWrapper, CcAgentWrapper, CodexAgentWrapper
+from reme.components.agent_wrapper.as_agent_wrapper import WorkspaceBackend
 from reme.components.agent_wrapper import base_agent_wrapper
+from reme.components.application_context import ApplicationContext
 from reme.components import base_component


@ -55,3 +58,50 @@ def test_init_logs_unknown_when_sdk_distribution_metadata_is_missing(monkeypatch
 def test_agent_wrappers_declare_sdk_package(wrapper_class, sdk_package):
    """Each concrete backend identifies the distribution that provides its SDK."""
    assert wrapper_class.SDK_PACKAGE == sdk_package
+
+
+def test_project_path_is_independent_from_runtime_workspace(tmp_path):
+    """Project assets can live outside the runtime workspace."""
+    workspace = tmp_path / "project" / ".reme"
+    wrapper = _VersionedAgentWrapper(
+        app_context=ApplicationContext(workspace_dir=str(workspace)),
+        project_path="..",
+    )
+
+    assert wrapper.workspace_path == workspace
+    assert wrapper.project_path == tmp_path / "project"
+    assert wrapper.cwd == tmp_path / "project"
+    assert wrapper.project_skills_root == tmp_path / "project" / "skills"
+
+
+@pytest.mark.parametrize("wrapper_class", [AsAgentWrapper, CcAgentWrapper, CodexAgentWrapper])
+def test_agent_wrappers_share_project_skill_resolution(tmp_path, wrapper_class):
+    """Every backend resolves selected skills through the base project root."""
+    workspace = tmp_path / "project" / ".reme"
+    skill = tmp_path / "project" / "skills" / "one"
+    skill.mkdir(parents=True)
+    (skill / "SKILL.md").write_text("# one", encoding="utf-8")
+    kwargs = {
+        "app_context": ApplicationContext(workspace_dir=str(workspace)),
+        "project_path": "..",
+    }
+    if wrapper_class is AsAgentWrapper:
+        kwargs["as_llm"] = ""
+    wrapper = wrapper_class(**kwargs)
+
+    assert wrapper._resolve_project_skills(["one", "one"]) == {"one": skill}  # pylint: disable=protected-access
+
+
+@pytest.mark.asyncio
+async def test_agentscope_backend_passes_configured_environment_to_bash(tmp_path, monkeypatch):
+    """AgentScope subprocesses receive config environment values explicitly."""
+    monkeypatch.setenv("REME_AGENT_ENV_TEST", "parent")
+    backend = WorkspaceBackend(str(tmp_path), {"REME_AGENT_ENV_TEST": "configured"})
+
+    result = await backend.exec_shell(
+        [sys.executable, "-c", "import os; print(os.environ['REME_AGENT_ENV_TEST'])"],
+        cwd=str(tmp_path),
+    )
+
+    assert result.exit_code == 0
+    assert result.stdout == b"configured\n"
--- a/tests/unit/test_cc_agent_wrapper.py
+++ b/tests/unit/test_cc_agent_wrapper.py
@ -34,6 +34,8 @@ def test_ensure_claude_skill_dir_adds_selected_skills_without_replacing_existing
    project_skills = tmp_path / "skills"
    (project_skills / "one").mkdir(parents=True)
    (project_skills / "two").mkdir()
+    for name in ("one", "two"):
+        (project_skills / name / "SKILL.md").write_text(f"# {name}", encoding="utf-8")
    config_dir = tmp_path / "mem_session" / "claude_config"

    for root in _skill_roots(tmp_path):
@ -55,6 +57,8 @@ def test_ensure_claude_skill_dir_all_adds_each_project_skill(tmp_path):
    project_skills = tmp_path / "skills"
    (project_skills / "one").mkdir(parents=True)
    (project_skills / "two").mkdir()
+    for name in ("one", "two"):
+        (project_skills / name / "SKILL.md").write_text(f"# {name}", encoding="utf-8")
    config_dir = tmp_path / "mem_session" / "claude_config"

    _wrapper(tmp_path)._ensure_claude_skill_dir(config_dir, "all")
@ -69,6 +73,7 @@ def test_ensure_claude_skill_dir_preserves_existing_directory_link(tmp_path):
    """An existing skills link is user-owned and remains untouched."""
    project_skills = tmp_path / "skills"
    (project_skills / "one").mkdir(parents=True)
+    (project_skills / "one" / "SKILL.md").write_text("# one", encoding="utf-8")
    config_dir = tmp_path / "mem_session" / "claude_config"
    legacy_root = tmp_path / ".claude" / "skills"
    legacy_root.parent.mkdir(parents=True)
@ -103,6 +108,8 @@ def test_configured_skills_use_latest_sdk_allowlist(tmp_path):
    project_skills = tmp_path / "skills"
    (project_skills / "one").mkdir(parents=True)
    (project_skills / "two").mkdir()
+    for name in ("one", "two"):
+        (project_skills / name / "SKILL.md").write_text(f"# {name}", encoding="utf-8")

    opts = _wrapper(tmp_path)._build_options("hello", skills=["one"])

@ -112,6 +119,25 @@ def test_configured_skills_use_latest_sdk_allowlist(tmp_path):
        assert not (root / "two").exists()


+def test_configured_project_path_sources_skills_outside_workspace(tmp_path):
+    """Claude Code links project skills while keeping sessions in the workspace."""
+    project = tmp_path / "project"
+    workspace = project / ".reme"
+    skill = project / "skills" / "serper-search"
+    skill.mkdir(parents=True)
+    (skill / "SKILL.md").write_text("# Serper Search", encoding="utf-8")
+    wrapper = CcAgentWrapper(
+        app_context=ApplicationContext(workspace_dir=str(workspace)),
+        project_path="..",
+    )
+
+    opts = wrapper._build_options("hello", skills=["serper-search"])
+
+    assert opts.cwd == project
+    assert (project / ".claude" / "skills" / "serper-search").resolve() == skill
+    assert (workspace / "mem_session" / "claude_config" / "skills" / "serper-search").resolve() == skill
+
+
 def test_sdk_native_system_prompt_preset_is_preserved(tmp_path):
    """System prompt dictionaries pass directly to the latest SDK."""
    opts = _wrapper(tmp_path)._build_options(
@ -130,6 +156,13 @@ def test_sdk_native_system_prompt_preset_is_preserved(tmp_path):
    }


+def test_web_search_is_disallowed_by_default(tmp_path):
+    """Claude Code keeps its normal tools except for web search."""
+    opts = _wrapper(tmp_path)._build_options("hello")
+
+    assert opts.disallowed_tools == ["WebSearch"]
+
+
 def test_api_credentials_use_only_wrapper_config(tmp_path, monkeypatch):
    """Claude Code credentials do not fall back to ambient or shared LLM configuration."""
    wrapper = _wrapper(tmp_path)
--- a/tests/unit/test_daily_paper.py
+++ b/tests/unit/test_daily_paper.py
@ -0,0 +1,515 @@
+"""Focused tests for the daily-paper cookbook workflow."""
+
+import datetime as dt
+import importlib
+import json
+from pathlib import Path
+import subprocess
+import sys
+from unittest.mock import MagicMock
+
+import frontmatter
+import httpx
+import pytest
+
+from reme.components import ApplicationContext
+from reme.components.agent_wrapper.base_agent_wrapper import BaseAgentWrapper
+from reme.components.runtime_context import RuntimeContext
+from reme.config.config_parser import _load_config
+from reme.schema import DailyBriefOutput, PaperInfo, PaperNoteOutput, PaperSelection
+from reme.steps.cookbook.daily_paper import (
+    DailyPaperAnalyzeStep,
+    DailyPaperCollectStep,
+    DailyPaperDigestStep,
+    DailyPaperRankStep,
+    DailyPaperSelectStep,
+)
+from reme.steps.cookbook.daily_paper import analyze, collect
+from reme.steps.cookbook.daily_paper.rank import build_candidate_pool, rrf_score
+from reme.steps.cookbook.dingtalk import DingTalkMarkdownSendStep
+from reme.steps.cookbook.dingtalk import send as dingtalk_send
+from reme.utils import arxiv as arxiv_utils
+from reme.utils.huggingface_papers import paper_ids_from_html, paper_info_from_payload
+
+
+class _QueuedAgentWrapper(BaseAgentWrapper):
+    """Return queued structured responses without contacting an LLM."""
+
+    def __init__(self, outputs: list[dict], **kwargs):
+        super().__init__(**kwargs)
+        self.outputs = list(outputs)
+        self.calls: list[dict] = []
+
+    async def reply(self, inputs, **kwargs) -> dict:
+        """Record the request and pop the next structured fixture."""
+        self.calls.append({"inputs": inputs, "kwargs": kwargs})
+        return {"structured_output": self.outputs.pop(0), "result": "ok"}
+
+
+def _paper(arxiv_id: str, *, title: str = "Paper", upvotes: int = 10) -> PaperInfo:
+    return PaperInfo(
+        arxiv_id=arxiv_id,
+        title=title,
+        summary=f"Summary for {title}",
+        authors=["A. Author"],
+        upvotes=upvotes,
+    )
+
+
+def test_hf_payload_and_html_normalization():
+    """HF list/detail shapes normalize and HTML rank order de-duplicates."""
+    payload = {
+        "paper": {
+            "id": "2607.16051",
+            "title": "Loop the Loopies!",
+            "summary": "Abstract",
+            "authors": [{"name": "Zitian Gao"}],
+            "upvotes": 53,
+            "githubRepo": "https://github.com/example/repo",
+        },
+        "organization": {"fullname": "IQuest"},
+    }
+
+    paper = paper_info_from_payload(payload)
+
+    assert paper.arxiv_id == "2607.16051"
+    assert paper.authors == ["Zitian Gao"]
+    assert paper.organization == "IQuest"
+    assert paper.github_repo == "https://github.com/example/repo"
+    assert paper_ids_from_html(
+        '<a href="/papers/2607.16051">one</a><a href="/papers/2607.16051">dup</a>'
+        '<a href="/papers/2607.10001">two</a>',
+    ) == ["2607.16051", "2607.10001"]
+
+
+def test_rrf_and_memory_candidate_reserve():
+    """RRF is exact and the candidate pool preserves a memory-related slot."""
+    general = _paper("2607.10001", title="General model", upvotes=100)
+    memory = _paper("2607.10002", title="Long-term memory for agents", upvotes=1)
+    general.fused_score = rrf_score(1, None, rrf_k=60, weekly_weight=0.7)
+    memory.fused_score = rrf_score(100, None, rrf_k=60, weekly_weight=0.7)
+
+    candidates = build_candidate_pool([general, memory], limit=2, memory_reserve=1)
+
+    assert candidates == [general, memory]
+    assert general.fused_score == pytest.approx(1 / 61)
+
+
+def test_history_exclusion_reads_prior_frontmatter_only(tmp_path: Path):
+    """Only prior dated paper notes contribute historical exclusions."""
+    prior = tmp_path / "daily" / "2026-07-20"
+    current = tmp_path / "daily" / "2026-07-21"
+    prior.mkdir(parents=True)
+    current.mkdir(parents=True)
+    (prior / "paper-2607.10001.md").write_text(
+        frontmatter.dumps(frontmatter.Post("body", arxiv_id="2607.10001")),
+        encoding="utf-8",
+    )
+    (current / "paper-2607.10002.md").write_text(
+        frontmatter.dumps(frontmatter.Post("body", arxiv_id="2607.10002")),
+        encoding="utf-8",
+    )
+
+    found = DailyPaperCollectStep.load_historical_arxiv_ids(
+        tmp_path,
+        dt.date(2026, 7, 21),
+        30,
+        "daily",
+    )
+
+    assert found == {"2607.10001"}
+
+
+@pytest.mark.asyncio
+async def test_arxiv_pdf_downloads_missing_cache_once(tmp_path: Path, monkeypatch):
+    """A missing PDF is downloaded atomically and then reused on the next lookup."""
+    requests: list[httpx.Request] = []
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        requests.append(request)
+        return httpx.Response(200, content=b"%PDF-downloaded")
+
+    transport = httpx.MockTransport(handler)
+    async_client = httpx.AsyncClient
+    monkeypatch.setattr(
+        arxiv_utils.httpx,
+        "AsyncClient",
+        lambda **kwargs: async_client(transport=transport, **kwargs),
+    )
+    target = tmp_path / "resource" / "papers" / "2607.10001.pdf"
+    client = arxiv_utils.ArxivPdfClient()
+
+    assert await client.download("2607.10001", target) == target
+    assert await client.download("2607.10001", target) == target
+
+    assert target.read_bytes() == b"%PDF-downloaded"
+    assert [str(request.url) for request in requests] == ["https://arxiv.org/pdf/2607.10001"]
+
+
+def test_standalone_config_uses_only_claude_code_and_eight_am_cron(monkeypatch):
+    """The standalone config schedules 08:00 and routes all agent work to CC."""
+    for name in (
+        "DINGTALK_APP_KEY",
+        "DINGTALK_APP_SECRET",
+        "DINGTALK_ROBOT_CODE",
+        "DINGTALK_CONVERSATION_IDS",
+    ):
+        monkeypatch.delenv(name, raising=False)
+    config = _load_config("daily_cookbook")
+
+    assert config.get("extends") is None
+    assert config["jobs"]["daily_paper_cron"]["cron"] == "0 8 * * *"
+    steps = config["jobs"]["daily_paper"]["steps"]
+    assert config["jobs"]["daily_paper_cron"]["steps"] == steps
+    agent_steps = [
+        step
+        for step in steps
+        if step["backend"]
+        in {
+            "daily_paper_select_step",
+            "daily_paper_analyze_step",
+            "daily_paper_digest_step",
+        }
+    ]
+    assert {step.get("agent_wrapper") for step in agent_steps} == {"claude_code"}
+    assert steps[-1] == {
+        "backend": "dingtalk_markdown_send_step",
+        "input_mapping": {"daily_paper_digest_path": "markdown_path"},
+        "app_key": "",
+        "app_secret": "",
+        "robot_code": "",
+        "conversation_ids": "",
+        "title": "ReMe Daily Paper",
+        "timeout": 15,
+    }
+    assert set(config["components"]["agent_wrapper"]) == {"claude_code"}
+    assert "as_llm" not in config["components"]
+    assert config["components"]["agent_wrapper"]["claude_code"]["project_path"] == ".."
+    assert "skills" not in config["components"]["agent_wrapper"]["claude_code"]
+
+
+def test_daily_paper_config_passes_dingtalk_environment(monkeypatch):
+    """The notifier receives all proactive-message settings from the environment."""
+    values = {
+        "DINGTALK_APP_KEY": "app-key",
+        "DINGTALK_APP_SECRET": "app-secret",
+        "DINGTALK_ROBOT_CODE": "robot-code",
+        "DINGTALK_CONVERSATION_IDS": "group-one,group-two",
+    }
+    for name, value in values.items():
+        monkeypatch.setenv(name, value)
+
+    step = _load_config("daily_cookbook")["jobs"]["daily_paper"]["steps"][-1]
+
+    assert {key: step[key] for key in ("app_key", "app_secret", "robot_code", "conversation_ids")} == {
+        "app_key": "app-key",
+        "app_secret": "app-secret",
+        "robot_code": "robot-code",
+        "conversation_ids": "group-one,group-two",
+    }
+
+
+def test_reme_import_does_not_require_optional_dingtalk_stream():
+    """Importing ReMe must not eagerly load the core-only DingTalk dependency."""
+    script = """
+import builtins
+
+original_import = builtins.__import__
+
+def guarded_import(name, *args, **kwargs):
+    if name == "dingtalk_stream":
+        raise ModuleNotFoundError("blocked optional dependency")
+    return original_import(name, *args, **kwargs)
+
+builtins.__import__ = guarded_import
+import reme
+"""
+
+    result = subprocess.run(
+        [sys.executable, "-c", script],
+        cwd=Path(__file__).parents[2],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    assert result.returncode == 0, result.stderr
+
+
+@pytest.mark.asyncio
+async def test_pipeline_filters_strict_yesterday_and_writes_outputs(
+    tmp_path: Path,
+    monkeypatch,
+):
+    """The complete mocked pipeline filters yesterday/history and writes linked notes."""
+    papers = {
+        "2607.10001": _paper("2607.10001", title="Best monthly paper", upvotes=100),
+        "2607.10002": _paper("2607.10002", title="Yesterday paper", upvotes=90),
+        "2607.10003": _paper("2607.10003", title="Previously recommended", upvotes=80),
+    }
+
+    prior_dir = tmp_path / "daily" / "2026-07-19"
+    prior_dir.mkdir(parents=True)
+    (prior_dir / "paper-2607.10003.md").write_text(
+        frontmatter.dumps(frontmatter.Post("old", arxiv_id="2607.10003")),
+        encoding="utf-8",
+    )
+
+    class _FakeHfClient:
+        requested_daily: list[str] = []
+
+        def __init__(self, **_kwargs):
+            pass
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, *_args):
+            return None
+
+        async def fetch_scope(self, scope: str, value: str):
+            """Return deterministic weekly/monthly fixtures."""
+            if scope == "month":
+                assert value == "2026-07"
+                return list(papers.values())
+            assert value == "2026-W30"
+            return [papers["2607.10001"], papers["2607.10002"]]
+
+        async def fetch_daily_ids(self, day: str):
+            """Record and return the exact requested day."""
+            self.requested_daily.append(day)
+            return {"2607.10002"}
+
+    async def fake_download(_self, _arxiv_id: str, target: Path):
+        """Create a minimal cached-PDF fixture."""
+        target.parent.mkdir(parents=True, exist_ok=True)
+        target.write_bytes(b"%PDF-fake")
+        return target
+
+    def fake_extract(_self, _path: Path, _max_pages: int, _max_chars: int):
+        """Return deterministic extracted text."""
+        return "--- PAGE 1 ---\nPaper content", 1, False
+
+    monkeypatch.setattr(collect, "HuggingFacePapersClient", _FakeHfClient)
+    monkeypatch.setattr(analyze.ArxivPdfClient, "download", fake_download)
+    monkeypatch.setattr(analyze.DailyPaperAnalyzeStep, "_extract_pdf_text_sync", fake_extract)
+
+    cc_wrapper = _QueuedAgentWrapper(
+        [
+            {
+                "selection_reasoning": "Best remaining ranked paper.",
+                "selected": [
+                    {
+                        "arxiv_id": "2607.10001",
+                        "rank": 1,
+                        "reason": "Strong result",
+                        "memory_relevance": "low",
+                    },
+                ],
+                "alternates": [],
+            },
+            {
+                "description": "Detailed note",
+                "body": "# Detailed reading\n\nEvidence [p. 1].",
+            },
+            {
+                "description": "Five-minute brief",
+                "body": "# 今日论文速读\n\n[[daily/2026-07-21/paper-2607.10001.md]]",
+            },
+        ],
+    )
+    app_context = ApplicationContext(
+        workspace_dir=str(tmp_path),
+        resource_dir="external-assets",
+        timezone="Asia/Shanghai",
+        language="zh",
+    )
+    context = RuntimeContext(
+        date="2026-07-21",
+        top_k=1,
+        candidate_limit=2,
+        memory_reserve=0,
+    )
+
+    await DailyPaperCollectStep(app_context=app_context)(context)
+    await DailyPaperRankStep(app_context=app_context)(context)
+    await DailyPaperSelectStep(app_context=app_context, agent_wrapper=cc_wrapper)(context)
+    await DailyPaperAnalyzeStep(app_context=app_context, agent_wrapper=cc_wrapper)(context)
+    await DailyPaperDigestStep(app_context=app_context, agent_wrapper=cc_wrapper)(context)
+
+    assert _FakeHfClient.requested_daily == ["2026-07-20"]
+    assert context.response.metadata["selected_arxiv_ids"] == ["2607.10001"]
+    assert context.response.metadata["excluded_yesterday_count"] == 1
+    assert context.response.metadata["excluded_history_count"] == 1
+    note_path = tmp_path / "daily" / "2026-07-21" / "paper-2607.10001.md"
+    digest_path = tmp_path / "daily" / "2026-07-21" / "daily-paper-brief.md"
+    note = frontmatter.load(note_path)
+    assert note.metadata["arxiv_id"] == "2607.10001"
+    assert note.metadata["source_pdf"] == "[[external-assets/papers/2607.10001.pdf]]"
+    assert (tmp_path / "external-assets" / "papers" / "2607.10001.pdf").is_file()
+    assert "[[daily/2026-07-21/paper-2607.10001.md]]" in digest_path.read_text(
+        encoding="utf-8",
+    )
+    assert not (tmp_path / "metadata" / "daily_paper" / "2026-07-21.json").exists()
+    digest = frontmatter.load(digest_path)
+    assert digest.metadata["selection_reasoning"] == "Best remaining ranked paper."
+    assert digest.metadata["arxiv_ids"] == ["2607.10001"]
+    assert all(set(call["kwargs"]) == {"output_schema"} for call in cc_wrapper.calls)
+    assert [call["kwargs"]["output_schema"] for call in cc_wrapper.calls] == [
+        PaperSelection,
+        PaperNoteOutput,
+        DailyBriefOutput,
+    ]
+    analysis_prompt = cc_wrapper.calls[1]["inputs"]
+    assert "长期记忆相关性初筛：low" in analysis_prompt
+    assert "必须先使用代码读取和搜索工具查看当前 ReMe 代码仓库" in analysis_prompt
+    assert "这应当是少数例外：一般情况下不要给建议" in analysis_prompt
+
+    rerun = RuntimeContext(date="2026-07-21")
+    await DailyPaperCollectStep(app_context=app_context)(rerun)
+    assert rerun.response.metadata["skipped"] is True
+    assert rerun.response.metadata["selection"] == {
+        "selection_reasoning": "Best remaining ranked paper.",
+        "selected": [
+            {
+                "arxiv_id": "2607.10001",
+                "rank": 1,
+                "reason": "Strong result",
+                "memory_relevance": "low",
+            },
+        ],
+        "alternates": [],
+    }
+    assert rerun.get("daily_paper_digest_path") == "daily/2026-07-21/daily-paper-brief.md"
+    assert _FakeHfClient.requested_daily == ["2026-07-20"]
+
+
+@pytest.mark.asyncio
+async def test_dingtalk_markdown_sends_groups_serially_in_configured_order(tmp_path: Path, monkeypatch):
+    """The notifier gets one app token and posts once per group in list order."""
+    digest_path = tmp_path / "daily" / "2026-07-21" / "daily-paper-brief.md"
+    digest_path.parent.mkdir(parents=True)
+    digest_path.write_text(
+        frontmatter.dumps(frontmatter.Post("# 今日论文\n\n测试内容", name="daily-paper-brief")),
+        encoding="utf-8",
+    )
+    token_calls = 0
+    seen_payloads: list[dict] = []
+
+    def get_access_token(client):
+        nonlocal token_calls
+        token_calls += 1
+        assert client.credential.client_id == "app-key"
+        assert client.credential.client_secret == "app-secret"
+        return "app-access-token"
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/v1.0/robot/groupMessages/send"
+        assert request.headers["x-acs-dingtalk-access-token"] == "app-access-token"
+        seen_payloads.append(json.loads(request.content))
+        return httpx.Response(200, json={"processQueryKey": f"query-{len(seen_payloads)}"})
+
+    transport = httpx.MockTransport(handler)
+    transport_kwargs: dict = {}
+
+    def ipv4_transport(**kwargs):
+        transport_kwargs.update(kwargs)
+        return transport
+
+    dingtalk_stream = importlib.import_module("dingtalk_stream")
+    monkeypatch.setattr(dingtalk_stream.DingTalkStreamClient, "get_access_token", get_access_token)
+    monkeypatch.setattr(dingtalk_send.httpx, "AsyncHTTPTransport", ipv4_transport)
+    app_context = ApplicationContext(workspace_dir=str(tmp_path))
+    context = RuntimeContext(markdown_path="daily/2026-07-21/daily-paper-brief.md")
+
+    step = DingTalkMarkdownSendStep(
+        app_context=app_context,
+        app_key="app-key",
+        app_secret="app-secret",
+        robot_code="robot-code",
+        conversation_ids=" group-one,group-two ",
+        title="ReMe Daily Paper",
+    )
+    step.logger = MagicMock()
+    response = await step(context)
+
+    assert token_calls == 1
+    assert transport_kwargs == {"local_address": "0.0.0.0"}
+    assert [payload["openConversationId"] for payload in seen_payloads] == ["group-one", "group-two"]
+    assert all(payload["robotCode"] == "robot-code" for payload in seen_payloads)
+    assert all(payload["msgKey"] == "sampleMarkdown" for payload in seen_payloads)
+    assert [json.loads(payload["msgParam"]) for payload in seen_payloads] == [
+        {"title": "ReMe Daily Paper", "text": "# 今日论文\n\n测试内容"},
+    ] * 2
+    assert response.metadata["dingtalk_configured_count"] == 2
+    assert response.metadata["dingtalk_sent_count"] == 2
+    logs = "\n".join(call.args[0] for call in step.logger.info.call_args_list)
+    assert "sending DingTalk Markdown" in logs
+    assert "delivery complete sent=2 total=2" in logs
+    assert all(value not in logs for value in ("app-key", "app-secret", "robot-code", "group-one", "group-two"))
+
+
+@pytest.mark.asyncio
+async def test_dingtalk_markdown_without_conversations_is_a_noop(tmp_path: Path):
+    """An empty conversation list keeps daily-paper generation usable without DingTalk."""
+    context = RuntimeContext(markdown_path="missing.md")
+
+    response = await DingTalkMarkdownSendStep(app_context=ApplicationContext(workspace_dir=str(tmp_path)))(context)
+
+    assert response.success is True
+    assert response.metadata["dingtalk_configured_count"] == 0
+    assert response.metadata["dingtalk_sent_count"] == 0
+
+
+@pytest.mark.asyncio
+async def test_existing_daily_paper_is_reused_and_sent_to_dingtalk(tmp_path: Path, monkeypatch):
+    """An idempotent daily-paper run skips generation but still notifies DingTalk."""
+    digest_path = tmp_path / "daily" / "2026-07-22" / "daily-paper-brief.md"
+    digest_path.parent.mkdir(parents=True)
+    digest_path.write_text(
+        frontmatter.dumps(frontmatter.Post("# 已有日报\n\n复用正文", name="daily-paper-brief")),
+        encoding="utf-8",
+    )
+    seen_payloads: list[dict] = []
+
+    dingtalk_stream = importlib.import_module("dingtalk_stream")
+    monkeypatch.setattr(
+        dingtalk_stream.DingTalkStreamClient,
+        "get_access_token",
+        lambda _client: "app-access-token",
+    )
+
+    async def handler(request: httpx.Request) -> httpx.Response:
+        seen_payloads.append(json.loads(request.content))
+        return httpx.Response(200, json={"processQueryKey": "query-1"})
+
+    transport = httpx.MockTransport(handler)
+
+    monkeypatch.setattr(dingtalk_send.httpx, "AsyncHTTPTransport", lambda **_kwargs: transport)
+    app_context = ApplicationContext(workspace_dir=str(tmp_path))
+    context = RuntimeContext(date="2026-07-22")
+
+    await DailyPaperCollectStep(app_context=app_context)(context)
+    response = await DingTalkMarkdownSendStep(
+        app_context=app_context,
+        input_mapping={"daily_paper_digest_path": "markdown_path"},
+        app_key="app-key",
+        app_secret="app-secret",
+        robot_code="robot-code",
+        conversation_ids="existing-group",
+        title="ReMe Daily Paper",
+    )(context)
+
+    assert response.metadata["skipped"] is True
+    assert response.metadata["dingtalk_sent_count"] == 1
+    assert seen_payloads == [
+        {
+            "robotCode": "robot-code",
+            "openConversationId": "existing-group",
+            "msgKey": "sampleMarkdown",
+            "msgParam": json.dumps(
+                {"title": "ReMe Daily Paper", "text": "# 已有日报\n\n复用正文"},
+                ensure_ascii=False,
+            ),
+        },
+    ]
--- a/tests/unit/test_dingtalk_wait.py
+++ b/tests/unit/test_dingtalk_wait.py
@ -0,0 +1,324 @@
+"""Focused tests for the DingTalk background agent bridge."""
+
+# pylint: disable=missing-function-docstring,protected-access
+
+import asyncio
+from types import SimpleNamespace
+from unittest.mock import MagicMock
+
+import pytest
+
+from reme.components import ApplicationContext, R
+from reme.components.agent_wrapper.base_agent_wrapper import BaseAgentWrapper
+from reme.config.config_parser import _load_config
+from reme.enumeration import ChunkEnum, ComponentEnum
+from reme.schema import StreamChunk
+from reme.steps.cookbook.dingtalk.wait import DingTalkWaitStep, _CardRenderer, _session_key
+
+
+class _AgentWrapper(BaseAgentWrapper):
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        self.calls = []
+
+    async def reply(self, inputs, **kwargs):
+        raise NotImplementedError
+
+    async def reply_stream(self, inputs, **kwargs):
+        self.calls.append((inputs, kwargs))
+        session_id = kwargs.get("resume") or "session-1"
+        yield StreamChunk(
+            chunk_type=ChunkEnum.THINK,
+            chunk="检查上下文",
+            block_id="think-1",
+            session_id=session_id,
+        )
+        yield StreamChunk(chunk_type=ChunkEnum.THINK, chunk="", block_id="think-1", session_id=session_id)
+        yield StreamChunk(chunk_type=ChunkEnum.CONTENT, chunk="回答", session_id=session_id)
+
+
+class _Card:
+    instances = []
+
+    def __init__(self, _client, _message):
+        self.card_template_id = "template"
+        self.title = None
+        self.markdown = ""
+        self.finished = False
+        self.failed = False
+        self.updates = []
+        self.stream_flags = []
+        self.content = ""
+        self.at_sender = False
+        self.__class__.instances.append(self)
+
+    def set_title_and_logo(self, title, _logo):
+        self.title = title
+
+    def get_card_data(self, flow_status=None):
+        data = {"msgContent": self.markdown}
+        if flow_status is not None:
+            data["flowStatus"] = flow_status
+        return data
+
+    async def async_create_and_send_card(self, _template, _data, at_sender=False):
+        self.at_sender = at_sender
+        return "card-1"
+
+    async def async_streaming(self, _card_id, _key, content, append, finished, failed):
+        self.updates.append(content)
+        self.stream_flags.append((append, finished, failed))
+        self.content = self.content + content if append else content
+        if finished:
+            self.finished = not failed
+            self.failed = failed
+
+
+class _Handler:
+    def __init__(self):
+        self.dingtalk_client = object()
+        self.replies = []
+
+    def reply_text(self, text, _message):
+        self.replies.append(text)
+
+
+class _WebSocket:
+    def __init__(self, messages=()):
+        self.messages = list(messages)
+        self.closed = asyncio.Event()
+
+    async def __aenter__(self):
+        return self
+
+    async def __aexit__(self, *_args):
+        await self.close()
+
+    def __aiter__(self):
+        return self
+
+    async def __anext__(self):
+        if self.messages:
+            return self.messages.pop(0)
+        await self.closed.wait()
+        raise StopAsyncIteration
+
+    async def close(self):
+        self.closed.set()
+
+
+class _StreamClient:
+    TAG_DISCONNECT = "disconnect"
+
+    def __init__(self, route_result=""):
+        self.route_result = route_result
+        self.websocket = None
+
+    def pre_start(self):
+        return None
+
+    def open_connection(self):
+        return {"endpoint": "wss://example.test/connect", "ticket": "ticket"}
+
+    async def keepalive(self, _websocket):
+        await asyncio.Event().wait()
+
+    async def route_message(self, _message):
+        return self.route_result
+
+
+def _message(text="hello", sender="user-1", conversation="cid-1", conversation_type="1"):
+    return SimpleNamespace(
+        text=SimpleNamespace(content=text),
+        sender_staff_id=sender,
+        conversation_id=conversation,
+        conversation_type=conversation_type,
+    )
+
+
+def test_session_key_uses_conversation_type_id_and_sender():
+    assert _session_key(_message()) == "1:cid-1:user-1"
+    assert _session_key(_message(sender="user-2")) == "1:cid-1:user-2"
+    assert _session_key(_message(conversation="cid-2", conversation_type="2")) == "2:cid-2:user-1"
+
+
+def test_renderer_keeps_blocks_in_stream_order_and_finishes_with_elapsed_time(monkeypatch):
+    now = 10.0
+    monkeypatch.setattr("reme.steps.cookbook.dingtalk.wait.time.monotonic", lambda: now)
+    renderer = _CardRenderer()
+    deltas = [
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.CONTENT, chunk="中间内容", block_id="content-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.CONTENT, chunk="", block_id="content-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.THINK, chunk="分析问题", block_id="think-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.THINK, chunk="", block_id="think-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.CONTENT, chunk="最终答案", block_id="answer-2")),
+    ]
+
+    markdown = "".join(deltas)
+    assert "### 💬 Content\n\n中间内容" in markdown
+    assert "### 🧠 Think\n\n分析问题" in markdown
+    assert "### 💬 Answer" not in markdown
+    assert markdown.count("### 💬 Content") == 2
+    assert markdown.index("中间内容") < markdown.index("### 🧠 Think") < markdown.index("最终答案")
+
+    now = 15.5
+    markdown += renderer.finish()
+    assert "### 💬 Answer" not in markdown
+    assert "### 💬 Content\n\n最终答案" in markdown
+    assert markdown.endswith("最终答案\n\n5.5s")
+    assert all(value not in markdown for value in ("ReMe Agent", "question", "正在生成", "执行轨迹"))
+    assert all(value not in markdown for value in ("✅", "### 当前"))
+
+
+def test_renderer_limits_non_content_blocks_to_100_characters():
+    renderer = _CardRenderer()
+    deltas = [
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.THINK, chunk="a" * 101, block_id="think-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.THINK, chunk="", block_id="think-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.TOOL_RESULT, chunk="b" * 101, block_id="tool-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.TOOL_RESULT, chunk="", block_id="tool-1")),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.CONTENT, chunk="c" * 101, block_id="answer-1")),
+        renderer.finish(),
+    ]
+    markdown = "".join(deltas)
+    for character in "ab":
+        assert f"{character * 100}..." in markdown
+        assert character * 101 not in markdown
+    assert "c" * 101 in markdown
+    assert f"{'c' * 100}..." not in markdown
+
+
+def test_renderer_keeps_tool_call_and_result_visible():
+    renderer = _CardRenderer()
+    tool_call_deltas = [
+        renderer.feed(
+            StreamChunk(
+                chunk_type=ChunkEnum.TOOL_CALL,
+                chunk='{"name":"Read","id":"tool-1"}',
+                block_id="metadata-1",
+                tool_call_name="Read",
+            ),
+        ),
+        renderer.feed(
+            StreamChunk(
+                chunk_type=ChunkEnum.TOOL_CALL,
+                chunk='{"path":',
+                block_id="tool-1",
+                tool_call_name="Read",
+            ),
+        ),
+        renderer.feed(
+            StreamChunk(chunk_type=ChunkEnum.TOOL_CALL, chunk='"memory.md"}', block_id="tool-1"),
+        ),
+        renderer.feed(StreamChunk(chunk_type=ChunkEnum.TOOL_CALL, chunk="", block_id="tool-1")),
+    ]
+    assert tool_call_deltas[:3] == ["", "", ""]
+
+    deltas = [
+        *tool_call_deltas,
+        renderer.feed(
+            StreamChunk(
+                chunk_type=ChunkEnum.TOOL_RESULT,
+                chunk={"content": '{"message":"memory content","count":1}'},
+                block_id="tool-1",
+            ),
+        ),
+        renderer.finish(),
+    ]
+    markdown = "".join(deltas)
+    assert markdown.count("### 🔧 Tool Call") == 1
+    assert "### 🔧 Tool Call · `Read`" in markdown
+    assert '> {\n> 　"path": "memory.md"\n> }' in markdown
+    assert "```" not in markdown
+    assert '"name": "Read"' not in markdown
+    assert "### 📦 Tool Result" in markdown
+    assert '> 　"content": {\n> 　　"message": "memory content",\n> 　　"count": 1\n> 　}' in markdown
+
+
+@pytest.mark.asyncio
+async def test_messages_resume_session_and_clear_only_the_combined_key(tmp_path):
+    app_context = ApplicationContext(workspace_dir=str(tmp_path))
+    wrapper = _AgentWrapper(app_context=app_context)
+    step = DingTalkWaitStep(app_context=app_context, agent_wrapper=wrapper, card_update_interval=0.05)
+    step.logger = MagicMock()
+    handler = _Handler()
+    sdk = SimpleNamespace(
+        AIMarkdownCardInstance=_Card,
+        AICardStatus=SimpleNamespace(PROCESSING="1"),
+    )
+    sessions = {}
+    message = _message()
+    key = _session_key(message)
+
+    await step._handle_message(message, key, sessions, handler, sdk)
+    await step._handle_message(message, key, sessions, handler, sdk)
+
+    assert sessions == {key: "session-1"}
+    assert wrapper.calls == [("hello", {}), ("hello", {"resume": "session-1"})]
+    assert all(card.finished for card in _Card.instances[-2:])
+    assert all(card.stream_flags for card in _Card.instances[-2:])
+    assert all(all(append for append, _finished, _failed in card.stream_flags) for card in _Card.instances[-2:])
+    assert all("### 💬 Content\n\n回答" in card.content for card in _Card.instances[-2:])
+    assert all(sum(update.count("检查上下文") for update in card.updates) == 1 for card in _Card.instances[-2:])
+    assert all(sum(update.count("回答") for update in card.updates) == 1 for card in _Card.instances[-2:])
+    assert all(card.at_sender for card in _Card.instances[-2:])
+    assert all(card.title is None for card in _Card.instances[-2:])
+
+    other_key = _session_key(_message(sender="user-2"))
+    sessions[other_key] = "session-2"
+    await step._handle_message(_message(text="/clear"), key, sessions, handler, sdk)
+    assert sessions == {other_key: "session-2"}
+    assert handler.replies[-1] == "✅ Conversation cleared. The next message will start a new session."
+    logs = "\n".join(call.args[0] for call in step.logger.info.call_args_list)
+    assert "received DingTalk text" in logs
+    assert "completed DingTalk reply" in logs
+    assert "cleared DingTalk session" in logs
+    assert "conversation_type='1' conversation_id='cid-1' sender_staff_id='user-1'" in logs
+    assert all(value not in logs for value in ("hello", "session-1"))
+
+
+def test_daily_cookbook_registers_one_step_background_wait_job(monkeypatch):
+    for name in ("DINGTALK_APP_KEY", "DINGTALK_APP_SECRET", "DINGTALK_ROBOT_CODE"):
+        monkeypatch.delenv(name, raising=False)
+    config = _load_config("daily_cookbook")
+    job = config["jobs"]["dingtalk_wait"]
+    assert job["backend"] == "background"
+    assert job["steps"] == [
+        {
+            "backend": "dingtalk_wait_step",
+            "agent_wrapper": "claude_code",
+            "app_key": "",
+            "app_secret": "",
+            "robot_code": "",
+            "card_update_interval": 1.0,
+            "worker_count": 4,
+        },
+    ]
+    assert R.get(ComponentEnum.STEP, "dingtalk_wait_step") is DingTalkWaitStep
+
+
+def test_daily_cookbook_passes_dingtalk_environment_to_step(monkeypatch):
+    monkeypatch.setenv("DINGTALK_APP_KEY", "app-key")
+    monkeypatch.setenv("DINGTALK_APP_SECRET", "app-secret")
+    monkeypatch.setenv("DINGTALK_ROBOT_CODE", "robot-code")
+    step = _load_config("daily_cookbook")["jobs"]["dingtalk_wait"]["steps"][0]
+    assert (step["app_key"], step["app_secret"], step["robot_code"]) == ("app-key", "app-secret", "robot-code")
+
+
+@pytest.mark.asyncio
+async def test_stream_client_closes_when_background_stop_is_set(monkeypatch):
+    websocket = _WebSocket()
+    monkeypatch.setattr("websockets.connect", lambda _uri: websocket)
+    stop_event = asyncio.Event()
+    task = asyncio.create_task(DingTalkWaitStep._run_client(_StreamClient(), stop_event))
+    stop_event.set()
+    await asyncio.wait_for(task, timeout=1)
+    assert websocket.closed.is_set()
+
+
+@pytest.mark.asyncio
+async def test_stream_client_restarts_after_server_disconnect(monkeypatch):
+    websocket = _WebSocket(["{}"])
+    monkeypatch.setattr("websockets.connect", lambda _uri: websocket)
+    with pytest.raises(ConnectionError, match="WebSocket closed"):
+        await DingTalkWaitStep._run_client(_StreamClient("disconnect"), asyncio.Event())