checkpoint

⚒️ Generated with [Fabro](https://fabro.sh)

stages/004-preflight_lint@1/output.log

@@ -0,0 +1 @@
+blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126

stages/004-preflight_lint@1/script_timing.json

@@ -0,0 +1,8 @@
+{
+  "output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
+  "exit_code": 0,
+  "duration_ms": 140689,
+  "termination": "exited",
+  "output_bytes": 0,
+  "live_streaming": false
+}

stages/004-preflight_lint@1/status.json

@@ -0,0 +1,6 @@
+{
+  "outcome": "succeeded",
+  "notes": "Script completed: cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
+  "failure_reason": null,
+  "timestamp": "2026-05-22T14:01:20.894716Z"
+}

stages/005-implement@1/prompt.md

@@ -0,0 +1,153 @@
+Goal: # Event-Sourced Agent Todo Tools
+
+Date: 2026-05-22
+
+## Summary
+
+Add one shared todo/task engine behind two model-native tool surfaces:
+
+- OpenAI models get Codex-compatible `update_plan`.
+- Anthropic models get Claude-compatible `TaskCreate`, `TaskUpdate`, and `TaskList`.
+- All mutations persist as individual `todo.created`, `todo.updated`, and `todo.deleted` run events.
+- `RunProjection` maintains current todo state by replaying those events.
+
+Scoping matches latest upstream behavior:
+
+- OpenAI plan todos are scoped to the emitting session: `openai_plan:<session_id>`.
+- Anthropic task todos are scoped to the root agent session: `anthropic_tasks:<root_session_id>`, shared by subagents.
+
+## Key Changes
+
+- Add shared todo domain types in `fabro-types`: `TodoStatus`, `TodoListKind`, `TodoProjection`, `TodoListProjection`, and `todos_by_list` on `RunProjection`.
+- Add run event bodies for `todo.created`, `todo.updated`, and `todo.deleted`; map them through Fabro's typed event pipeline and replay them in `fabro-store`'s `RunProjectionReducer`.
+- Extend `fabro-agent` tool runtime context with `session_id`, `root_session_id`, `tool_call_id`, and a narrow agent-event emitter so tools can emit todo mutation events with correct session metadata.
+- Add a shared `TodoRuntime` in `fabro-agent` that owns the in-memory current todo projection for active sessions and emits individual mutation events.
+- Register `update_plan` only in `OpenAiProfile`. It accepts Codex's schema, diffs incoming steps by exact `step`, and emits create/update/delete events so the projected list equals the submitted plan.
+- Register `TaskCreate`, `TaskUpdate`, and `TaskList` only in `AnthropicProfile`. Use generated numeric task IDs per Anthropic task list, preserve Claude field names and result text, and support status `deleted` as a delete operation.
+- Thread root session identity through parent and child sessions. Root sessions use their own ID as `root_session_id`; subagent sessions inherit the parent root ID while retaining their own `session_id`.
+- Update OpenAPI `RunProjection` schema and regenerate Rust/TypeScript API types. Existing run-state and run-events APIs remain the exposure point; no new HTTP route is required.
+- Update web run-event invalidation so `todo.*` events refresh `getRunState` consumers and the run events list.
+
+## Tool Semantics
+
+### OpenAI `update_plan`
+
+Input:
+
+- `explanation?: string`
+- `plan: [{ step: string, status: "pending" | "in_progress" | "completed" }]`
+
+Behavior:
+
+- Return Codex-compatible success text: `Plan updated`.
+- Scope todos to `openai_plan:<session_id>`.
+- Use exact `step` string as identity within that scope.
+- Reject duplicate `step` strings with a model-visible tool error.
+- New step string emits `todo.created`.
+- Existing step with changed status or order emits `todo.updated`.
+- Omitted previous step emits `todo.deleted`.
+- Todo ID is deterministic from `list_id + step`.
+
+### Anthropic `TaskCreate`
+
+Input:
+
+- `subject: string`
+- `description: string`
+- `activeForm?: string`
+- `metadata?: object`
+
+Behavior:
+
+- Scope tasks to `anthropic_tasks:<root_session_id>`.
+- Generate numeric task IDs per Anthropic task list.
+- Emit `todo.created`.
+- Return `Task #<id> created successfully: <subject>`.
+
+### Anthropic `TaskUpdate`
+
+Input:
+
+- `taskId: string`
+- `subject?: string`
+- `description?: string`
+- `activeForm?: string`
+- `status?: "pending" | "in_progress" | "completed" | "deleted"`
+- `owner?: string`
+- `addBlocks?: string[]`
+- `addBlockedBy?: string[]`
+- `metadata?: object`
+
+Behavior:
+
+- `status: "deleted"` emits `todo.deleted`.
+- Other changes emit `todo.updated`.
+- Metadata merges into existing metadata; a `null` metadata value deletes that key.
+- Missing task returns a non-error tool result: `Task not found`.
+
+### Anthropic `TaskList`
+
+Input:
+
+```json
+{}
+```
+
+Behavior:
+
+- Reads the projected Anthropic task list for `anthropic_tasks:<root_session_id>`.
+- Returns Claude-style task lines containing ID, status, subject, optional owner, and uncompleted blockers.
+- Returns `No tasks found` when the list is empty.
+
+## Test Plan
+
+- Unit-test tool schemas and registration:
+  - OpenAI profile includes `update_plan`; Anthropic profile does not.
+  - Anthropic profile includes `TaskCreate`, `TaskUpdate`, `TaskList`; OpenAI profile does not.
+- Unit-test OpenAI reconciliation:
+  - create initial steps
+  - update status and order
+  - delete omitted steps
+  - subagent session writes to a different OpenAI list than parent
+- Unit-test Anthropic task behavior:
+  - create, list, update, complete, and delete
+  - metadata merge and `null` deletion
+  - `addBlocks` and `addBlockedBy`
+  - parent and subagent share the same Anthropic task list
+- Unit-test event serialization and conversion:
+  - `todo.created`, `todo.updated`, `todo.deleted` have stable event names and properties
+  - envelope includes emitting `session_id`, `parent_session_id` when present, and `tool_call_id`
+- Unit-test `RunProjection` replay:
+  - replaying todo events reconstructs current lists
+  - deleted todos are absent from current projection
+  - multiple lists in one run remain isolated
+- API and web checks:
+  - `cargo build -p fabro-api`
+  - `cargo nextest run -p fabro-server`
+  - `cd lib/packages/fabro-api-client && bun run generate`
+  - `cd apps/fabro-web && bun test && bun run typecheck`
+
+## Assumptions
+
+- Ignore deprecated Claude `TodoWrite`; implement only latest Anthropic Task tools.
+- Duplicate OpenAI `step` strings are rejected with a model-visible tool error because step text is the identity.
+- Todo events are individual mutations only; no plan-replace event is added.
+- Todo state is exposed through existing run projection and event APIs, not a new endpoint.
+
+
+## Completed stages
+- **toolchain**: succeeded
+  - Script: `command -v cargo >/dev/null || { curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && sudo ln -sf $HOME/.cargo/bin/* /usr/local/bin/; }; cargo --version 2>&1`
+  - Output:
+    ```
+    cargo 1.95.0 (f2d3ce0bd 2026-03-21)
+    ```
+- **preflight_compile**: succeeded
+  - Script: `cargo check -q --workspace 2>&1`
+  - Output: (empty)
+- **preflight_lint**: succeeded
+  - Script: `cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1`
+  - Output: (empty)
+
+
+Read the plan file referenced in the goal and implement every step. Make all the code changes described in the plan. Use red/green TDD.

stages/005-implement@1/provider_used.json

@@ -0,0 +1,5 @@
+{
+  "mode": "agent",
+  "provider": "anthropic",
+  "model": "claude-opus-4-7"
+}

stages/005-implement@1/response.md

@@ -0,0 +1,48 @@
+The implementation is complete. Summary of changes:
+
+## Summary
+
+**Domain types in `fabro-types`** (new `todo.rs`):
+- `TodoStatus` (pending/in_progress/completed/deleted), `TodoListKind` (openai_plan/anthropic_tasks), `TodoProjection`, `TodoListProjection` with `upsert`/`remove`/`get`.
+- Added `todos_by_list: BTreeMap<String, TodoListProjection>` to `RunProjection`.
+
+**Run-event bodies** (new `run_event/todo.rs`):
+- `TodoCreatedProps`, `TodoUpdatedProps`, `TodoDeletedProps` with full field-level patch semantics for updates.
+- Wired into `EventBody` enum, `event_name()` match, and `is_known_event_name`.
+
+**Reducer in `fabro-store`**:
+- Three new arms in `RunProjectionReducer::apply_event` that replay `todo.created`/`updated`/`deleted` into `todos_by_list`. Supports metadata patches with `null` deletion and `addBlocks`/`addBlockedBy` deduplication.
+
+**Agent runtime (`fabro-agent`)**:
+- `AgentEvent::TodoCreated/Updated/Deleted` variants with full payloads.
+- Extended `ToolContext` with `session_id`, `root_session_id`, `tool_call_id`, and an `AgentEventEmitter` trait.
+- New `TodoRuntime` (`todo_runtime.rs`) maintaining in-memory projection per `list_id`, emitting one event per mutation.
+- New `update_plan` (OpenAI), `TaskCreate`/`TaskUpdate`/`TaskList` (Anthropic) tools in `todo_tools.rs`, registered only in their respective profiles.
+- `Session.root_session_id` field with accessor + setter; `spawn_agent` tool inherits the parent's root into subagent sessions.
+- `Emitter::emit_with_parent` + `SessionBoundEmitter` plug into `tool_execution::execute_one_tool`, populating `ToolContext` with the live session metadata.
+
+**Conversion in `fabro-workflow`**:
+- Added arms in `event/convert.rs` and `event/names.rs` to translate `AgentEvent::Todo*` into `EventBody::Todo*`.
+
+**OpenAPI + TypeScript**:
+- Extended `docs/public/api-reference/fabro-api.yaml` with `todos_by_list`, `TodoStatus`, `TodoListKind`, `TodoProjection`, `TodoListProjection` schemas.
+- Added corresponding TS models in `lib/packages/fabro-api-client/src/models/` and updated `RunProjection` model.
+
+**Web (`apps/fabro-web`)**:
+- Added `TODO_EVENTS` set in `run-events.ts`; `todo.*` events now invalidate `queryKeys.runs.state(id)` and `queryKeys.runs.events(id, 1000)`, plus the current stage's events.
+
+**Tests added** (all passing):
+- 4 unit tests in `fabro-types::todo` (`list_id_is_prefix_colon_session`, ordering, upsert/remove).
+- 4 event-serde tests in `fabro-types::run_event` (event names known, envelope metadata, round-trip create/delete).
+- 4 reducer tests in `fabro-store` (replay reconstructs list, deletes absent, multi-list isolation, metadata patch + null).
+- 3 `TodoRuntime` tests (create/update/delete cycle, deleted-status emits delete, missing-todo returns false).
+- 11 tool tests in `todo_tools` (`update_plan`: create initial / status+order / delete omitted / reject duplicate / subagent different list; `TaskCreate`/`TaskUpdate`/`TaskList`: numeric IDs, cycle, metadata merge + null, addBlocks dedupe, empty list, missing taskId, parent+subagent share Anthropic list).
+- Profile tests updated to assert exact tool registration (OpenAI has `update_plan`, Anthropic has the three Task tools, neither bleeds).
+- 2 web tests for `queryKeysForRunEvent` on `todo.*` events.
+
+**Verified clean**:
+- `cargo +nightly-2026-04-14 fmt --check --all` — clean.
+- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings` — clean.
+- `cargo nextest run` across `fabro-types`, `fabro-store`, `fabro-workflow`, `fabro-agent`, `fabro-server`, `fabro-api`: 2638 passed (3 pre-existing graphviz-env failures unrelated to this change).
+- `cargo build -p fabro-api` — clean.
+- `cd apps/fabro-web && bun run typecheck` — clean. `bun test`: 421 pass (7 pre-existing failures unrelated to this change).