mirror of
https://github.com/alirezarezvani/claude-skills.git
synced 2026-10-10 03:27:56 +00:00
890 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Claude
|
d36ec93d3b
|
chore(v2.7.0): register 12 v2 skills in marketplace + codex sync
Resolves the two follow-ups left after the v2 megaprompt sweep. **marketplace.json (.claude-plugin/):** +12 plugin entries for the new v2 skills across 3 categories. Categories added: productivity (3), research (8). | Plugin | Source | Category | |---|---|---| | capture-skill | ./productivity/capture | productivity | | email-pair | ./productivity/email | productivity | | reflect-skill | ./productivity/reflect | productivity | | landing | ./marketing/landing | marketing | | pulse | ./research/pulse | research | | litreview | ./research/litreview | research | | grants | ./research/grants | research | | dossier | ./research/dossier | research | | patent | ./research/patent | research | | syllabus | ./research/syllabus | research | | notebooklm | ./research/notebooklm | research | | research-orchestrator | ./research/research | research | Per CLAUDE.md ClawHub rules: cs- prefix not used in repo registry (reserved for ClawHub when slug conflicts arise). 12 entries cover 13 skills (email-pair holds inbox-setup + inbox-triage). Total plugins in marketplace: 43 → 55. **.codex sync:** Manually created 11 missing symlinks under .codex/skills/ (capture + pulse already existed from prior auto-sync). Added 13 entries to .codex/skills-index.json with category metadata. Total skills: 290 → 303. **scripts/sync-codex-skills.py:** Added productivity/, marketing/, and research/ to SKILL_DOMAINS so future automated sync runs pick up the new top-level domains (previously only domain folders were registered). Manual symlink creation deliberately avoids the script's full --dry-run-flagged 16 [UPDATED] symlinks on pre-existing duplicate-path skills (chief-ai-officer-advisor, chaos-engineering, etc., which have both nested-plugin and flat paths in the repo). That churn belongs in a separate cleanup PR — not in this release-prep PR. Validation: - marketplace.json: 55 plugins, all 7 required fields, no duplicates - skills-index.json: 303 entries across 11 categories - all 13 megaprompt symlinks resolve correctly - 8-phase plugin audit on research/research: PASS WITH WARNINGS (Phase 2 structure 84.1/GOOD, Phase 5 security PASS, scripts 3/3) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
c63822c579
|
Merge pull request #669 from alirezarezvani/claude/build-skills-notebooklm-browser-auto | ||
|
Alireza Rezvani
|
575805aa99
|
Merge pull request #671 from alirezarezvani/claude/skills-library-megaprompts-c1sQy | ||
|
Claude
|
4803bb8180
|
feat(research): orchestrator — Path-B hybrid router + fallback from megaprompt 13
Slice 7 (final v2 megaprompt). Architecture C: deterministic SIGNALS classification → specialist delegation (≥2 signals OR single weak match) OR own 8-step plan-decompose-search-synthesize-cite fallback. Routing transparency is mandatory — never delegates silently. Always states the decision + accepts override. Override is logged. Distinct from engineering/autoresearch-agent (Karpathy's file-optimization loop) — completely different use case. README + plugin.json + SKILL.md all call out the disambiguation explicitly. After this merges: ALL 13 v2 megaprompts shipped. 11 files, 1,659 lines: - .claude-plugin/plugin.json (with distinct_from autoresearch-agent) - README.md (disambiguation table + routing target table) - agents/cs-research.md (router persona, routing-transparency enforcer) - commands/cs-research.md (/cs:research <question>) - skills/research/SKILL.md (full Path-B converted spec) - skills/research/references/hybrid_router_architecture.md (8 sources) - skills/research/references/deterministic_classification_canon.md (7 sources) - skills/research/references/fallback_workflow_canon.md (7 sources) - skills/research/scripts/classifier.py (stdlib, SIGNALS map + scoring) - skills/research/scripts/routing_transparency_logger.py (stdlib, JSON audit) - skills/research/scripts/fallback_decomposer.py (stdlib, 3-5 sub-questions) All 3 scripts smoke-tested: - classifier --sample → litreview routed (3 signals: pico + systematic review + meta-analysis) - classifier "research microsoft" → fallback (0 signals — correct, generic "research X" must not auto-route) - classifier "FTO landscape" → patent routed (weak: 1 signal, single specialist) - logger --sample → 4-event sequence (decision → delegation → decision → override) persisted to ~/.research_sessions/sample.json - decomposer --sample → 5 sub-questions via what/why/how/who/what's next framework Path-B fidelity: SIGNALS map preserved verbatim from post-PR-#657 audit (no bracketed placeholders; verb-noun pairs only). All anti-patterns from the megaprompt encoded in SKILL.md. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
ac6db2fab7
|
feat(research): notebooklm — Path-B browser-automation slice from megaprompt 03
Slice 6: browser-automation shape — the only such skill in the v2 collection. Distinct from research-pack convention (no Agent Integrity Rules, no 1 q/sec, no DOCX). Action-routing intake (Q1 picks one of 4 actions). After this merges: 11 of 13 v2 megaprompts shipped. Only Slice 7 (13-research orchestrator) remains. SOURCE SPEC megaprompts/03-notebooklm-megaprompt.md (PR #657). WHAT THE SKILL DOES Controls Google NotebookLM via browser automation. 4 core actions: 1. Read/Extract — chat-based extraction from existing notebook 2. Add Sources — push URL/text/file/Google Doc/synthesized content 3. Studio Outputs — 9 types (Audio Overview, Study Guide, Briefing Doc, Timeline, FAQ, Table of Contents, Infographic, Slides, Mind Map) with MANDATORY custom prompts 4. Create New Notebook — initialize with title + initial sources DOMAIN FOLDER research/. Semantic domain is research (users automate NotebookLM as part of their research workflow). But technical shape is completely different from research-pack siblings — distinct enough that the README + SKILL.md explicitly call out the shape difference. KEY PATH-B PRESERVED ELEMENTS - Critical portability notice at top — requires browser automation; graceful failure in non-automation contexts (Step 0 check) - Action-routing intake: Q1 forces 1-of-4 action commitment; refuses to start without it - Q2-Q4 branch per action (per-source-type for Q3, mandatory custom prompt for Q4 when Studio) - Screenshot-first discipline (NotebookLM is dynamic SPA) - find()-before-click semantic finder discipline - Tool-agnostic vocabulary (no "Claude Chrome Extension" hardcoding) - Never auto-handle login (detect login wall → halt, never type credentials) - Async fire-and-notify pattern for slow Studio ops (Audio Overview 5-10 min, Infographic/Slides/Mind Map 2-5 min) - Studio customization menu MANDATORY (chevron, not main button — defaults produce mediocre output) - File upload via file-upload tool, NOT native picker REPO STRUCTURE research/notebooklm/ ├── .claude-plugin/plugin.json ├── README.md ├── agents/cs-notebooklm.md ← browser-automation persona, │ async-discipline + screenshot │ enforcer ├── commands/cs-notebooklm.md ← /cs:notebooklm └── skills/notebooklm/ ├── SKILL.md ├── references/ │ ├── browser_automation_canon.md ← screenshot-first + │ │ find-before-click + │ │ tool-agnostic (7 sources: │ │ Anthropic Computer Use, │ │ Playwright, Selenium, │ │ WebDriver, MS Power Auto, │ │ ARIA, Anthropic cookbook) │ ├── studio_output_custom_prompts.md ← per-output-type templates │ │ (7 sources: NotebookLM │ │ docs, Anthropic prompt │ │ eng, Refactoring UI, │ │ Gallo Talk Like TED, │ │ Lencioni BLUF, Bloom, │ │ Tufte) │ └── async_action_discipline.md ← fire-and-notify canon │ (7 sources: Anthropic API │ timeouts, NotebookLM │ timing data, Erlang let- │ it-crash, AWS Step │ Functions, Twelve-Factor, │ Node event loop, │ Playwright/Selenium) └── scripts/ ├── action_router.py ← stdlib: Q1-Q4 → action │ plan + UI flow + required │ params + per-source-type │ and per-studio-type │ branching + validation ├── custom_prompt_template_generator.py ← stdlib: output type + │ audience + length + angle │ → starter prompt for 9 │ studio output types └── async_action_classifier.py ← stdlib: action → WAIT (with timeout) or FIRE_AND_NOTIFY (with notify message) 11 files, 2,003 lines. VERIFIED CLEAN All 3 scripts pass smoke tests: - action_router: sample (studio + audio_overview + custom prompt) → correctly identifies FIRE_AND_NOTIFY timing, lists 11-step UI flow, flags critical rule "ALWAYS open customization menu (chevron) — NEVER click main Studio button". Add-source URL action → 3 screenshots, WAIT timing. Validation: studio without ≥30-char custom prompt → FAIL with explicit error. - custom_prompt_template_generator: sample (audio_overview + executive + compact) → produces complete starter prompt naming audience role, length, focus, structural requirements. Study_guide + undergraduate variant correctly applies "Define every technical term. Assume zero specialized background" rule. - async_action_classifier: audio_overview → FIRE_AND_NOTIFY (5-10 min, with notify message template). chat_send → WAIT (3-10s, 30s timeout, 3s polling). infographic → FIRE_AND_NOTIFY (2-5 min). All 16 documented actions routable. All 3 with --output json: valid JSON. plugin.json validates. VERTICAL-SLICE STATUS ✓ Slice 1: capture (PR #659) ✓ Slice 2: pulse (PR #660) ✓ Slice 3: email pair (PR #661) ✓ Slice 4: landing (PR #662) ✓ Slice 5 batch 1: litreview (PR #663) ✓ Slice 5 batch 2: grants + dossier (PR #664) ✓ Slice 5 batch 3: patent + syllabus (PR #666) ✓ Cleanup PR: move pulse + capture (PR #667) ✓ Slice 8: reflect (PR #668) ✓ Slice 6: notebooklm (this PR) ☐ Slice 7: 13-research orchestrator + autoresearch-agent reconciliation 11 of 13 v2 megaprompts shipped after this merge. Only Slice 7 remains, then v2 is complete. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json not updated (separate concern; done after all 13 ship) - .codex/skills/notebooklm symlink not added (auto-sync workflow handles on merge) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
f2cd5f2dda
|
Merge pull request #668 from alirezarezvani/claude/build-skills-reflect-productivity | ||
|
Claude
|
7bdc98e517
|
feat(productivity): reflect skill — Path-B light-prompt-flow sibling of capture
Slice 8: productivity light-prompt-flow sibling. Same shape as capture (11 files, max-1-question intake, fast-to-action), different mode — capture organizes external dumps; reflect re-examines internal conversation state. After this merges: 10 of 13 v2 megaprompts shipped. SOURCE SPEC megaprompts/02-reflect-megaprompt.md (PR #657). WHAT THE SKILL DOES Mid-conversation reflection. Pauses execution, re-reads the FULL conversation from original goal forward (not just recent turns), runs the 5-dimension analysis framework: - Macro Perspective (original goal vs current; drift detection) - Gap Analysis (assumptions / stakeholders / constraints / alternatives / external factors) - Reflective Inquiry (right problem? simpler path? harder valuable path avoided?) - Bias Check (confirmation / sunk cost / anchoring / complexity / recency — each with recognition cues) - Contextual Alignment (does direction serve actual goals + best use of time + external factors) Delivers flowing prose (NO headers, NO bullets). Ends with mandatory directional recommendation: Continue / Pivot to {X} / Pause for {Q}. KEY PATH-B PRESERVED ELEMENTS - Re-read FULL conversation from original goal (not just recent turns) — the discipline that distinguishes real reflection from local summary - Halt-current-thread stop directive (reflection is a pause, not a side-quest) - Honest-output discipline: NO manufactured problems when path is solid; NO vague reassurance ("looks good!") instead of specific reasoning - 5-dimension framework preserved verbatim - 5 biases preserved (confirmation, sunk cost, anchoring, complexity, recency) with recognition cues - Flowing prose enforced (no headers, no bullets in body) - Closing recommendation mandatory (Continue / Pivot to X / Pause for Q) - Low-intake: max 1 optional clarifier (only when context is thin); default to no questions - No name references (generic second-person throughout) - Implicit triggers OFFER reflection, never auto-invoke (10+ detail turns / frustration / dead-ends → ask user if they want to step back, don't unilaterally run) PURE-REASONING SKILL No external APIs. No DOCX generation. No file-system writes beyond audit. Most portable v2 skill — works in Claude Code CLI + Claude.ai web natively, no MCP dependencies, no Node.js, no Consensus account required. REPO STRUCTURE (mirrors capture 1:1) productivity/reflect/ ├── .claude-plugin/plugin.json ├── README.md ├── agents/cs-reflect.md ← reflection persona, honest-output enforcer ├── commands/cs-reflect.md ← /cs:reflect (or auto-triggers on phrases) └── skills/reflect/ ├── SKILL.md ├── references/ │ ├── cognitive_bias_canon.md ← 5 biases + recognition cues │ (7 sources: Tversky/Kahneman, │ Wason, Arkes/Blumer, │ Russo/Schoemaker, Tetlock, │ Karpathy) │ ├── honest_output_discipline.md ← anti-manufactured-problems │ (7 sources: Yegge, Gawande, │ Deming, Russell, Kim Scott, │ Bret Victor, skill spec) │ └── conversation_reflection_practice.md ← Schön reflective practice │ (7 sources: Schön 1983 + 1987, │ Argyris/Schön, Kolb, Polanyi, │ Kahneman/Tversky, Victor) └── scripts/ ├── bias_pattern_detector.py ← stdlib: regex scan for 5-bias │ signal patterns ├── conversation_depth_analyzer.py ← stdlib: turn count + implicit │ trigger signal detection └── directional_recommendation_validator.py ← stdlib: verify output ends with Continue/Pivot/Pause + specific evidence + flowing prose 11 files, 1,554 lines. Comparable to capture (1,560 lines). VERIFIED CLEAN All 3 scripts pass smoke tests: - bias_pattern_detector --sample (notification system + sunk cost + anchoring + complexity scenario): correctly detects 3 biases (sunk_cost via "we've invested", anchoring via "sticking with", complexity via 9 "what about X" hits). Correctly clears confirmation + recency (no strong signals). - conversation_depth_analyzer --sample (19-turn debugging conversation with stuck-ness markers): correctly verdicts OFFER_REFLECT based on frustration (8 hits) + dead-ends (4 hits). Note: "skill should OFFER reflection, not auto-invoke" — honors design intent. - directional_recommendation_validator --sample-pass (honest validation output with 16 specific-evidence references): PASS 6/6. - directional_recommendation_validator --sample-fail (vague reassurance with bullets, no recommendation, no evidence): FAIL with 4 specific issues caught (missing closing recommendation, 2 vague phrases, 3 bullets, 0 specific-evidence refs). All 3 with --output json: valid JSON. plugin.json validates. VERTICAL-SLICE STATUS ✓ Slice 1: capture (PR #659) ✓ Slice 2: pulse (PR #660) ✓ Slice 3: email pair (PR #661) ✓ Slice 4: landing (PR #662) ✓ Slice 5 batch 1: litreview (PR #663) ✓ Slice 5 batch 2: grants + dossier (PR #664) ✓ Slice 5 batch 3: patent + syllabus (PR #666) ✓ Cleanup PR: move pulse + capture (PR #667) ✓ Slice 8: reflect (this PR) ☐ Slice 6: notebooklm (browser-automation, last shape) ☐ Slice 7: 13-research orchestrator + autoresearch-agent reconciliation 10 of 13 v2 megaprompts shipped after this merge. 3 remaining: notebooklm (browser-automation), 13-research (orchestrator), then v2 is complete. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json not updated (separate concern; done after all 13 ship) - .codex/skills/reflect symlink not added (auto-sync workflow handles on merge per existing pattern) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
alirezarezvani
|
bc487ee041 |
chore: sync codex skills symlinks [automated]
Some checks are pending
Sync Codex Skills Symlinks / sync (push) Waiting to run
|
||
|
Alireza Rezvani
|
7ebbb52cc2
|
Merge pull request #667 from alirezarezvani/claude/cleanup-move-pulse-capture | ||
|
Claude
|
6d9630f83c
|
chore(cleanup): move pulse + capture to proper domain folders
Surgical move PR — resolves the two domain warts accumulated during
the v2 megaprompt build sweep:
engineering/pulse/ → research/pulse/ (research-pack — pulse is
the first research skill;
now joins litreview, grants,
dossier, patent, syllabus)
engineering/capture/ → productivity/capture/ (productivity — capture
is brain-dump organizer,
not engineering tooling)
WHY THIS PR
When Slice 1 (capture) shipped in PR #659, the productivity/ domain
folder didn't yet exist. When Slice 2 (pulse) shipped in PR #660, the
research/ folder didn't yet exist either. Both were placed in
engineering/ as the catch-all.
After Slices 3-5 established the productivity/, marketing/, and
research/ top-level domain folders, those two early skills were left
in engineering/ as warts. This PR resolves them BEFORE Slice 7
(13-research orchestrator) so the orchestrator can reference
research/pulse/ as its routing target without further path churn.
WHAT MOVED
Two directories moved via `git mv` (preserves rename history):
- engineering/pulse → research/pulse (11 files)
- engineering/capture → productivity/capture (11 files)
INTERNAL REFERENCES UPDATED
Inside the moved directories:
- .claude-plugin/plugin.json homepage URLs (engineering/X → new path)
- agents/cs-*.md `skills:` frontmatter field
CROSS-SKILL REFERENCES UPDATED
6 external files reference pulse and/or capture as sibling skills.
All updated via sed:
productivity/email/agents/cs-inbox-setup.md (capture ref)
productivity/email/agents/cs-inbox-triage.md (pulse + capture refs)
research/grants/agents/cs-grants.md (pulse ref)
research/litreview/agents/cs-litreview.md (pulse ref + stale
"will move in cleanup
PR" caveat removed)
research/dossier/agents/cs-dossier.md (pulse ref)
marketing/landing/agents/cs-landing.md (pulse + capture refs)
CODEX SYMLINKS RE-POINTED
.codex/skills/{capture,pulse} symlinks updated to point at new
locations. Verified resolution to SKILL.md files works.
.codex/skills-index.json still references the old paths — this file
is auto-regenerated by the codex-sync workflow on every merge to dev
(prior commits:
|
||
|
Alireza Rezvani
|
f0176e0bd9
|
Merge pull request #666 from alirezarezvani/claude/build-skills-research-batch-3 | ||
|
Claude
|
50b5b1b9ed
|
feat(research): patent + syllabus — Path-B batch 3 (specialty research-pack variants)
Slice 5 batch 3 — final two research-pack siblings. Specialty variants: - patent: 5-sub-use-case routing (novelty/FTO/landscape/diligence/litigation) - syllabus: BUNDLED-JS-DOCX-GENERATOR pattern (first in repo) After this merges: ALL 6 research-pack siblings shipped (pulse + litreview + grants + dossier + patent + syllabus). 9 of 13 v2 megaprompts complete. SOURCE SPECS - megaprompts/11-patent-megaprompt.md (PR #657) - megaprompts/10-syllabus-megaprompt.md (PR #657) PATENT (Prior-Art + Landscape Intelligence) Refuses generic "patent help". Q2 forces commitment to ONE of 5 sub-use-cases: novelty → narrow + claim-text focused; verdict NOVEL/POTENTIALLY/NOT NOVEL FTO → active patents only, jurisdiction-filtered; CLEAR/FLAGGED/HIGH RISK per jurisdiction landscape → CPC trends + filer tally; CONCENTRATED/COMPETITIVE/EMERGING diligence → assignee + assignment chain + family resolution; PORTFOLIO VERIFIED/PARTIAL/RISK litigation → adjacent art before priority date; KNOCK-OUT/STRONG/WEAK/NO MATERIAL ART Each sub-use-case uses fundamentally different search strategy (enforced by sub_use_case_router.py). DOCX section emphasis varies per sub-use-case. Key Path-B preserved elements: - 6-Q grill-me intake with Q2 mandatory commitment + Q3-Q6 conditional skips - 4 sources: Google Patents (workhorse) + Espacenet + USPTO + Lens.org BYOK - CPC/IPC class follow-up after initial keyword search (catches keyword-missed art) - Family resolution across jurisdictions (deduplicates same-invention filings) - Date discipline (filing/priority/publication/grant — surface legally-relevant) - Mandatory legal disclaimer for novelty + FTO (Q6 triggers) - Out-of-scope flagging (trademark/copyright/trade-secret) - 8-section DOCX with sub-use-case-specific emphasis Scripts: - citation_tracker.py: multi-source three-count (Google Patents + Espacenet + USPTO + Lens.org) + 1s sequential discipline + Lens BYOK tracking - family_resolver.py: 3-pass clustering (family_id → priority_number → heuristic with 80% Jaccard on assignee + inventor + matching priority_date) - sub_use_case_router.py: deterministic strategy from 5 sub-use-cases → query plan + ranking heuristic + DOCX emphasis flags + legal disclaimer flag References (7+ sources each): - sub_use_case_routing.md: MPEP, 35 USC 102/103, WIPO PCT, EPO Guidelines, USPTO PPS docs, Google Patents docs, Lens.org API - cpc_classification_canon.md: CPC scheme, WIPO IPC, Mowery/Nelson/Sampat, WIPO PATENTSCOPE, Cohen/Nelson/Walsh, MPEP §901, Lemley/Sampat - legal_disclaimer_discipline.md: MPEP §1.4-§1.5, AIPLA Code of Ethics, 35 USC §282/§271, EPO Guidelines, PCT Article 39, Fischer/Henkel on PAEs SYLLABUS (Course Supplementary Reading List) Bundled-JS variant — generates .docx via scripts/generate_reading_list.js (Node.js + docx package, ~395 lines) rather than inlining 300+ lines of DOCX layout in SKILL.md. Key Path-B preserved elements: - 3-Q grill-me intake (input format + audience + year range) - Group-and-confirm checkpoint after Phase 2 (proceed/merge/split/add/remove) - Applied-domain weaving (e.g., "enzyme kinetics food processing" not just "enzyme kinetics" — boosts relevance dramatically) - Audience calibration (undergrad-intro defines every term; grad-doctoral assumes technical fluency) - Bloom higher-order discussion questions (apply/analyze/evaluate, NOT recall) - Sequential Consensus 1 q/sec - Source discipline + three-count tracking - Bundled JS for DOCX (token-efficient + reusable + maintainable) Scripts: - citation_tracker.py: Consensus three-count + per-section breakdown + 1s sequential discipline - topic_grouper.py: greedy clustering of extracted topics into 6-12 sections via shared-keyword detection (≥2 significant words shared → same section); auto-merge smallest if >12, auto-split largest if <6 - discussion_question_validator.py: Bloom-level classification per question; flags BELOW-audience FAIL with verb-replacement suggestions; flags ABOVE-audience WARN - generate_reading_list.js: BUNDLED Node.js DOCX generator (~395 lines). Multi-location require fallback for `docx` package. JSON input → .docx output. Title page + intro + learning outcomes box + numbered papers per section + audit log + footer. References (7+ sources each): - applied_domain_weaving.md: Bloom 1956, Mayer multimedia learning, Fink significant learning, Donald disciplinary thinking, Lave/Wenger situated learning, Chickering/Gamson 7 principles, Boyer scholarship of application - audience_calibration.md: Bloom/Anderson-Krathwohl revised taxonomy, Marzano new taxonomy, Hattie visible learning, Bain great teachers, Walvoord/Anderson effective grading, Brookfield/Preskill discussion, Bjork desirable difficulty - bundled_script_pattern.md: Karpathy-coder discipline, CLAUDE.md anti-patterns, docx Node.js package, CommonJS module resolution, Twelve-Factor App, Kernighan/Plauger Software Tools, McIlroy/Unix philosophy REPO STRUCTURE Both plugins in research/. Patent uses standard 11-file layout. Syllabus uses 12-file layout (extra file: scripts/generate_reading_list.js bundled JS). VERIFIED CLEAN All 7 scripts pass smoke tests: Patent: - sub_use_case_router: FTO with US+EP → 8 queries with jurisdiction scaling; novelty (no jurisdictions) → 6 queries with claim-focused ranking - family_resolver: 6 sample hits → correctly resolves to 3 unique families (Acme: 3 jurisdictions; Beta: 2 jurisdictions; Gamma: 1). Deduplication savings: 3 - citation_tracker: lifecycle works, multi-source counts (Google Patents + Espacenet + USPTO + Lens) tracked separately, audit block matches DOCX Section 8 format Syllabus: - topic_grouper: 19 sample topics → 12 sections, headings derived from shared keywords ("Plant + Physiology", "Animal + Anatomy", etc.) - discussion_question_validator: 5 sample questions correctly classified. "What did authors find?" → recall, OK for undergrad_intro, FAIL for grad_doctoral with verb-replacement suggestions. "Design a follow-up study..." → create, OK for grad_doctoral. - citation_tracker: lifecycle works with per-section breakdown - generate_reading_list.js: syntax valid (node --check passes) All scripts: --output json valid. plugin.json validates. VERTICAL-SLICE STATUS ✓ Slice 1: capture (PR #659) ✓ Slice 2: pulse (PR #660) ✓ Slice 3: email pair (PR #661) ✓ Slice 4: landing (PR #662) ✓ Slice 5 batch 1: litreview (PR #663) ✓ Slice 5 batch 2: grants + dossier (PR #664) ✓ Slice 5 batch 3: patent + syllabus (this PR) ☐ Slice 6: notebooklm (browser-automation, last shape) ☐ Slice 7: 13-research orchestrator + autoresearch-agent reconciliation ☐ Slice 8: 02-reflect (productivity) ☐ Cleanup PR: move engineering/pulse + engineering/capture 9 of 13 skills shipped after this merge. ALL research-pack siblings complete (6 of 6 in research/). Only 3 special-shape skills remain: notebooklm (browser-auto), 13-research (orchestrator), 02-reflect (productivity). NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json: separate concern - .codex/skills/ symlinks: auto-sync on merge - engineering/pulse + engineering/capture: cleanup PR queued https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
0bb96856ea
|
Merge pull request #664 from alirezarezvani/claude/build-skills-research-batch-2 | ||
|
Claude
|
24a0bcd9f7
|
feat(research): grants + dossier — Path-B batch 2 (research-pack siblings)
Slice 5 batch 2: two research-pack siblings from megaprompts 08 + 12. Same shape as litreview (Slice 5 batch 1) with domain-specific variants: - grants: multi-source (Consensus + RePORTER POST + NOSI), 9-section DOCX - dossier: hypothesis-testing variant (Q4 mandatory, ≥30% disconfirming rule enforced), source-tier discipline (primary/secondary/tertiary) SOURCE SPECS - megaprompts/08-grants-megaprompt.md - megaprompts/12-dossier-megaprompt.md (Both PR #657. Canonical specs.) GRANTS (NIH Funding Intelligence) For clinical researchers — 6-Q grill-me (research idea + career stage + prelim + environment + posture + institutes) → 5-facet Consensus positioning → RePORTER POST institute mapping → NOSI fetches → 9-section .docx with MANDATORY program officer recommendation. Key Path-B preserved elements: - RePORTER POST-only constraint (web_fetch is GET — must use bash_tool + curl). Documented prominently in SKILL.md + reference + command. - Dynamic fiscal year computation (Oct 1 = new FY). - Scope-aware mechanism matching (NOT career stage alone — common failure mode). - Mandatory program officer recommendation (single highest-leverage pre-submission step). - Plan-tier detection from Consensus "Found N, showing top M" pattern. - 9 DOCX sections including Audit Log. Scripts: - citation_tracker.py: multi-source three-count audit (Consensus sent/shown/cited + RePORTER projects/cited + NOSI fetches with success/total) + 1s sequential discipline enforcement - fiscal_year_calculator.py: Oct-boundary-aware FY computation, no hardcoded years - mechanism_matcher.py: 3D lookup (career × scope × prelim) with environment override (R15 for resource-constrained), warnings for common mismatches References (7+ sources each): - nih_mechanism_matching.md: Sackett, Rockey, Robertson, NIH RePORTER, Mehrotra, NRSA guidelines, Heggeness - reporter_post_patterns.md: RePORTER API v2 docs, NIH Guide for Grants, praw etiquette, Cohen backoff, curl docs, Maynez on hallucinated citations, Susskind audit-log - docx_9_sections.md: docx lib, NIH OER writing strategies, Russell & Morrison Grant Writers' Workbook, PRISMA, RePORTER, Heggeness, Strunk & White DOSSIER (Decision-Grade Entity Research) Hypothesis-testing variant — refuses to be "tell me about Microsoft". Q4 (your hypothesis) is MANDATORY; ≥30% of search budget allocated to disconfirming queries. Source-tier discipline (primary/secondary/ tertiary) on every flag. Key Path-B preserved elements: - Non-generic framing prominently in SKILL.md ("the forcing Q4 is what makes this skill non-generic") - Q4 mandatory with implicit-fallback flag if user refuses after one push-back - ≥30% disconfirming rule documented + enforced via stdlib tool - Subject-type routing (person/company/nonprofit/gov source matrices) - Source-tier on every flag in DOCX - 9 DOCX sections including verdict (SUPPORTED/PARTIALLY/DISPROVEN/ INCONCLUSIVE) - Conversation hooks finding-tied, not generic - BYOK MCP usage flagged in audit log - Sensitivity exclusions (Q6) honored Scripts: - citation_tracker.py: three-count + supporting/disconfirming classification per query + source-tier per citation + tier-weighted verdict computation + BYOK MCP tracking - disconfirming_evidence_balance.py: enforces ≥30% rule with PASS/WARN/FAIL verdicts + antonym-pivot suggestions for adding disconfirming queries (antonym pivots like consolidating → diversifying, growing → shrinking, hiring → laying off) - source_tier_classifier.py: URL → tier via comprehensive domain pattern matching (SEC/court/.gov primary, NYT/WSJ/TechCrunch secondary, Reddit/HN/Glassdoor tertiary, blog hosting platforms pattern-matched, company-official heuristic via subject keywords) References (7+ sources each): - hypothesis_testing_discipline.md: Popper Logic of Scientific Discovery, Kahneman, Tetlock Superforecasting, Dawes, Taleb Black Swan, Popper Conjectures & Refutations, Levitin - subject_type_source_matrix.md: SEC EDGAR docs, ProPublica Nonprofit Explorer, FIPS/open-data, Pickering progressive enhancement, Schneier provenance, OWASP, Charity Navigator - conversation_hook_quality.md: Carnegie, Cialdini, Voss calibrated questions, Goleman EI, Lencioni trust, Gallo TED rhetoric, Schein humble inquiry REPO STRUCTURE Both plugins in research/ (the new domain folder from Slice 5 batch 1). Mirrors litreview's structure exactly: plugin.json + README + agents/cs-* + commands/cs-* + skills/<name>/SKILL.md + 3 refs + 3 scripts = 11 files per skill, 22 total. VERIFIED CLEAN All 6 scripts pass smoke tests: Grants: - fiscal_year_calculator: Oct 2026 → FY 2027 ✓; Sep 2026 → FY 2026 ✓ - mechanism_matcher: --sample (early career + pilot) returns 4 K-series mechanisms with full rationale + budget + best-for. With resource-constrained env, correctly leads with R15 (the targeted mechanism). - citation_tracker: lifecycle works. Sequential discipline enforced. Multi-source counts (Consensus + RePORTER + NOSI) correctly aggregated. Audit-block output matches DOCX Section 9 format. Dossier: - source_tier_classifier: 11 sample URLs correctly tiered (SEC/Microsoft official/ProPublica/Scholar/FederalRegister = primary; NYT/TechCrunch = secondary; HN/Glassdoor/Medium = tertiary; unknown blog = secondary with low-confidence note). - disconfirming_evidence_balance: sample (80% supporting / 20% disconfirming) correctly returns WARN with antonym-pivot suggestions ("consolidating → diversifying"). FAIL threshold triggers at <20%. - citation_tracker: lifecycle works. Supporting/disconfirming classification tracked. Source-tier per citation. Verdict computed (INCONCLUSIVE on <3 cited). BYOK MCP usage tracked. All 6 scripts: --output json valid. plugin.json validates. VERTICAL-SLICE STATUS ✓ Slice 1: capture (light prompt-flow, PR #659) ✓ Slice 2: pulse (research-pack, PR #660) ✓ Slice 3: email pair (workflow-pair, PR #661) ✓ Slice 4: landing (generator, PR #662) ✓ Slice 5 batch 1: litreview (academic research, PR #663) ✓ Slice 5 batch 2: grants + dossier (this PR) ☐ Slice 5 batch 3: patent + syllabus (specialty variants) ☐ Slice 6: notebooklm (browser-automation) ☐ Slice 7: 13-research orchestrator + autoresearch-agent reconciliation ☐ Slice 8: 02-reflect (productivity) ☐ Cleanup PR: move engineering/pulse + engineering/capture 7 of 13 skills shipped after this merge. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json: separate concern, after all 13 ship - .codex/skills/ symlinks: auto-sync workflow on merge - engineering/pulse + engineering/capture: cleanup PR queued https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
71a683e00d
|
Merge pull request #663 from alirezarezvani/claude/build-skills-research-batch-1 | ||
|
Claude
|
a4bb1fc648
|
feat(research): litreview skill — Path-B research-pack sibling from megaprompt 09
Slice 5 batch 1 of N: first research-pack sibling after pulse (Slice 2). Establishes the academic-literature variant of the research-pack shape + introduces the research/ top-level domain folder. SOURCE SPEC megaprompts/09-litreview-megaprompt.md (PR #657). Canonical. DOMAIN FOLDER DECISION (research/ — new) Pulse currently lives in engineering/ (placed before the domain-folder discipline crystallized). The right home for academic-research skills is research/ — parallel to productivity/, marketing/. This PR creates that folder; pulse + capture moves are deferred to a coordinated cleanup PR. Cumulative folder warts: - engineering/capture/ → productivity/capture/ (Slice 1 wart) - engineering/pulse/ → research/pulse/ (Slice 2 wart) Cleanup PR will address both before Slice 5 (orchestrator) lands. WHY ONE SKILL PER PR (NOT BATCH 6) User recommended batching 6 research-pack siblings. Doing 1 in this PR instead, with rationale: each megaprompt is dense (litreview alone is 266 lines with 8 DOCX sections, 3-tier search budget logic, cross-search intelligence trackers). 33-66 files of unfocused conversion risks Path-B fidelity. Subsequent PRs will ratchet up to 2 skills each now that the academic-literature pattern is validated. WHAT THE SKILL DOES Turns a research question into a strategically planned mini literature review delivered as an 8-section .docx. Grill-me intake (question + framework + tentative depth) before reconnaissance; second forcing checkpoint after Phase 2 confirms framework + sub-areas + final depth. Sequential Consensus searches at 1 q/sec, budget-allocated by tier (5/10/20). Cross-search intelligence (repeat-hits, recurring-authors, citations-per-year) feeds the "Start Here" + "Key Research Groups" DOCX sections. Output is a "launching pad" — orientation guide, not a finished review. PATH-B FIDELITY (megaprompt → SKILL.md) - Frontmatter description preserved verbatim from megaprompt. - 10-step workflow structure preserved 1:1 (Agent Integrity Rules → Error Handling → Phase 0 intake → Phase 1 recon → Phase 2 framework → Checkpoint → Phase 3 searches → Phase 4 DOCX → Doc structure → Technical requirements). - All 3 grill-me intake questions preserved verbatim with rationale. - All 5 Agent Integrity Rules preserved verbatim per PR #657 audit. - All 3 search budget tiers fully allocated (5/10/20 with explicit query breakdown per tier). - All 8 DOCX sections fully specified. - Interactive checkpoint described as forcing-options moment (not free-text). - Three frameworks (PICO/SPIDER/Decomposition) + Hybrid documented with examples. - Anti-patterns + error-handling table + validation checklist preserved. RESEARCH-PACK CONVENTION MARKERS (in SKILL.md per PR #657 audit) Agent Integrity Rules: 2 sequential: 5 three-count: 1 plan-tier: 3 1 query/sec: 2 checkpoint: 9 retry once: 2 Source discipline: 1 3 consecutive: 2 All markers present multiple times. REPO STRUCTURE research/litreview/ ├── .claude-plugin/plugin.json ← source.spec → megaprompts/09 ├── README.md ├── agents/cs-litreview.md ← sequential-Consensus + checkpoint enforcer ├── commands/cs-litreview.md ← /cs:litreview <research-question> └── skills/litreview/ ├── SKILL.md ← Path-B converted ├── references/ │ ├── framework_selection.md ← PICO/SPIDER/Decomp/Hybrid + 7 sources │ │ (Sackett, Cooke et al., Booth, PRISMA, │ │ Cochrane, Hewitt-Taylor, JBI) │ ├── search_budget_allocation.md ← 5/10/20 + cross-search + 7 sources │ │ (Consensus docs, Cochrane, Greenhalgh, │ │ PRISMA, Sandelowski, Lawani, AWS) │ └── docx_8_sections.md ← 8-section guide + 7 sources (docx lib, │ OOXML, PRISMA, Cochrane, Lipsey, Tufte, │ Strunk) └── scripts/ ├── citation_tracker.py ← stdlib: three-count + 1s rate-limit │ discipline enforcement ├── framework_recommender.py ← stdlib: keyword heuristic PICO/SPIDER/ │ Decomp/Hybrid recommendation └── cross_search_aggregator.py ← stdlib: repeat-hits, recurring-authors, citations-per-year ranking 11 files, 2,020 lines. Slightly heavier than pulse (1,643) due to: - Denser SKILL.md (251 lines vs pulse 258 — comparable) - Heaviest reference: docx_8_sections.md at 287 lines (8 sections × ~35 lines each spec) - citation_tracker.py is heavier than pulse's (258 vs 251) because it enforces the 1s sequential gap explicitly VERIFIED CLEAN - citation_tracker.py: full lifecycle works. Sequential discipline enforced — second search at 0.04s correctly REJECTED with "wait 0.96s more"; after 1.1s sleep, accepted. Three-count audit block output matches PR #657 audit format. - framework_recommender.py: PICO question (clinical reasoning vs physicians) → recommends PICO SPIDER question (qualitative burnout study) → recommends SPIDER with high confidence (3 signals) Decomposition question (RAG systems benchmarks) → recommends Decomposition (after plural-aware regex fix; was originally missed due to "systems" not matching "system") - cross_search_aggregator.py: sample with overlapping searches: Repeat-hits: Med-PaLM benchmark correctly flagged (3 sub-areas) Recurring authors: Singhal correctly top (4 appearances) Citations-per-year: USMLE benchmark paper top at 266/yr - All 3 with --output json: valid JSON - plugin.json validates; conforms to repo schema NOT-YET-DONE (for upcoming PRs) - Slice 5 batch 2: grants + dossier (next PR; same shape as litreview) - Slice 5 batch 3: patent + syllabus (specialty variants — patent has sub-use-case routing, syllabus has bundled JS DOCX generator) - Slice 6: notebooklm (browser-automation shape, separate slice) - Slice 7: 13-research orchestrator + autoresearch-agent reconciliation - Slice 8: 02-reflect productivity - Cleanup PR: move engineering/pulse + engineering/capture to their proper domain folders https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
2fa288e835
|
Merge pull request #662 from alirezarezvani/claude/build-skills-landing-generator-slice | ||
|
Claude
|
8690081a04
|
feat(marketing): landing skill — Path-B generator slice from megaprompt 04
Slice 4 of 13: generator shape. Validates the Path-B conversion pattern for skills that produce a single artifact (HTML file) with motion-design discipline. Also introduces the marketing/ domain folder (parallel to productivity/, separate from the existing structured marketing-skill/ folder that houses the 44 pod-based marketing skills). SOURCE SPEC megaprompts/04-landing-megaprompt.md (PR #657). The megaprompt is the canonical spec; this plugin is the working implementation. WHAT THE SKILL DOES Premium single-file HTML landing page generator. Outputs one polished .html file with GSAP 3D animations, scroll-triggered reveals, and mouse-parallax depth. All CSS inline, all JS inline; only externals are Google Fonts (Inter) + GSAP via CDN. Phase 0: 4 forcing intake questions (one at a time): Q1 — product/service (refuses vague pitches) Q2 — audience register (technical/business/consumer/internal) Q3 — brand overrides (HEX vars, or "default") Q4 — tone (professional/playful/authoritative/minimal) Then generates a single .html with three sections (Hero, Features, Closing CTA), GSAP entrance timeline, mouse-parallax handler, ScrollTrigger feature reveals, CSS floating shapes, scroll indicator. DOMAIN FOLDER DECISION (marketing/ — new) The existing marketing-skill/ folder houses 44 pod-based marketing skills with its own internal structure. v2 megaprompt-derived landing skill is a single self-contained plugin — placing it inside marketing-skill/ would disrupt that folder's pod structure. Creating marketing/ alongside it, parallel to the new productivity/ folder, gives the v2 megaprompt slices a clean visually-parallel home. Both folders coexist. DISAMBIGUATION FROM EXISTING landing-page-generator The repo already has product-team/skills/landing-page-generator/ — the v2.1.2 work that outputs Next.js TSX + Tailwind for conversion-optimized lead-gen with copy frameworks (PAS/AIDA/BAB). The v2 megaprompt 04 is a DIFFERENT skill: single-file HTML with GSAP for premium visual one-pagers. Different output (HTML vs TSX), different optimization target (visual premium vs conversion), different animation approach (GSAP vs static). Both skills coexist. README.md disambiguates clearly. Pick by use case: visual premium one-pager → marketing/landing/ conversion lead-gen → product-team/skills/landing-page-generator/ PATH-B CONVERSION DISCIPLINE - Frontmatter description preserved verbatim from megaprompt spec. - Workflow structure (megaprompt lines 28-43) became SKILL.md section ordering 1:1. - All 4 grill-me intake questions preserved verbatim with "why I'm asking" rationale. - All 5 animation patterns preserved (Hero Entrance / Mouse Parallax / ScrollTrigger Reveals / CSS Floats / Scroll Indicator). - Default brand palette preserved verbatim (--navy / --teal / --teal-glow / --amber / --off-white / --text-muted / --card-bg / --card-border). - All 3 sections (Hero, Features, Closing CTA) preserved with full spec. - Required CDN dependencies preserved verbatim. - Anti-patterns + error-handling table + validation checklist preserved. REPO STRUCTURE marketing/landing/ ├── .claude-plugin/plugin.json ← source.spec field points at megaprompt ├── README.md ← disambig from landing-page-generator ├── agents/cs-landing.md ← landing generator persona, FOUC enforcer ├── commands/cs-landing.md ← /cs:landing └── skills/landing/ ├── SKILL.md ← Path-B converted from megaprompt 04 ├── references/ │ ├── brand_system_design.md ← color theory + WCAG + algorithmic │ │ derivation (7 sources: WCAG 2.2, │ │ Refactoring UI, Material Design, │ │ IBM Carbon, APCA, Tailwind, etc.) │ ├── gsap_animation_patterns.md ← 5 animation patterns canon (7 sources: │ │ GSAP docs, Val Head, Rachel Nabors, │ │ Sarah Drasner, GPU-accel CSS, Material │ │ motion, WCAG 2.3.3) │ └── single_file_html_discipline.md ← inline + CDN-only rationale (7 sources: │ MDN, Inclusive Components, Resilient │ Web Design, no-build advocacy, etc.) └── scripts/ ├── brand_palette_validator.py ← stdlib: HEX validation + WCAG contrast │ + algorithmic palette derivation in HSL ├── kebab_slug_generator.py ← stdlib: name → kebab + duplicate detection └── html_validator.py ← stdlib: 11-rule structural post-gen check 11 files, 1,979 lines. Slightly heavier than capture (1,560) and pulse (1,643) due to denser SKILL.md (346 lines — full CSS + JS code blocks for all 5 animation patterns) and heavier html_validator.py (11 rules vs the simpler 7-rule checks in other slices). VERIFIED CLEAN - brand_palette_validator.py: sample (orange primary + teal accent + near-black bg) correctly surfaces real WCAG issue: white text on #FF6B35 = 2.64:1 (FAILs body-text 4.5:1 threshold). Real-world validation working as designed. Algorithmic palette derivation produces full var set in HSL space. - kebab_slug_generator.py: "Quill AI — Async Standup Tool" → kebab slug correctly handling em-dash. Duplicate detection + timestamp suffix suggestion. - html_validator.py: 17/17 PASS on clean sample; correctly catches 9 FAILs + 6 WARNs on violation sample (missing viewport, external CSS file, external JS file, missing 2 of 3 sections, missing gsap.set() before timeline, missing both responsive breakpoints, div with onclick, duplicate H1). - All 3 with --output json: valid JSON. - plugin.json validates; conforms to repo schema with source + distinct_from attribution. MEGAPROMPT FIDELITY MARKERS (in SKILL.md) Phase 0: 1 Hero: 18 900px: 3 gsap.set: 6 Features: 5 580px: 3 mouse parallax: 3 Closing CTA: 1 OUTPUT_DIR: 3 ScrollTrigger: 4 GSAP: 25 kebab: 4 CSS keyframes: 1 Google Fonts: 5 FOUC: 3 All megaprompt-mandated terms surface multiple times. VERTICAL-SLICE STATUS ✓ Slice 1: capture (light prompt-flow, PR #659 merged) ✓ Slice 2: pulse (research-pack, PR #660 merged) ✓ Slice 3: email-pair (workflow-pair, PR #661 merged) ✓ Slice 4: landing (generator — this PR; introduces marketing/) ☐ Slice 5: orchestrator/router (13-research) — last shape; must reconcile with existing engineering/autoresearch-agent/ After Slice 5, all 5 shapes are validated. The remaining 8 megaprompts (02-reflect light prompt-flow + 6 research-pack siblings + 03-notebooklm which is research-flavored) can be batched in larger PRs grouped by shape. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json not updated (separate concern; done after all 13 ship) - .codex/skills/landing symlink not added (auto-sync workflow handles on merge per existing pattern) - engineering/capture/ NOT moved to productivity/capture/ (would break anyone who installed from current path; address in separate cleanup PR) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
6483349a7a
|
Merge pull request #661 from alirezarezvani/claude/build-skills-email-workflow-pair | ||
|
Claude
|
a2e9e48eb2
|
feat(productivity): email pair (inbox-setup + inbox-triage) — Path-B workflow-pair slice
Slice 3 of 13: workflow-pair shape. Validates the Path-B conversion pattern for two coupled skills sharing a strict 7-file KB contract. Also introduces the productivity/ domain folder per the navigation-map distinction in CLAUDE.md (engineering/ = software-engineering scope; productivity/ = generic productivity workflows). DOMAIN FOLDER DECISION CLAUDE.md defines engineering/ as "Engineering (POWERFUL) — Agent design, RAG, MCP, CI/CD, database, observability." Email triage is generic productivity, not software engineering. New folder: productivity/. Capture (Slice 1, merged in PR #659) was placed under engineering/ before this distinction was sharpened. It will move to productivity/ in a separate cleanup PR — moving a merged plugin in this slice would risk breaking anyone who installed it from engineering/capture/. Future productivity slices (02-reflect) will go under productivity/ from the start. SOURCE SPECS - megaprompts/06-inbox-setup-megaprompt.md (PR #657) - megaprompts/07-inbox-triage-megaprompt.md (PR #657) The megaprompts are canonical; these plugins are working implementations. PR #657's cross-skill consistency audit verified the 7 KB filenames align verbatim between the two megaprompts. This slice preserves that alignment. WHAT THE PAIR DOES Two coupled skills sharing a 7-file KB at ${WORKSPACE}/Email/: inbox-setup (run once): Interactive 8-section interview (~25-31 grill-me questions) → writes 7 KB files (taxonomy, patterns, evaluation-framework, rate-card, blocklist, tracker, triage-log/). inbox-triage (run recurringly): Light-intake (max 2 optional override questions). Reads 7 KB files, classifies recent emails, researches new senders, generates recommendations (TAKE IT / WORTH / PASS / FLAG), drafts replies (NEVER SENDS), delivers report, updates blocklist + tracker, writes per-run log. 10 execution steps. PATH-B CONVERSION DISCIPLINE - Both megaprompts' frontmatter descriptions preserved verbatim. - Both workflow structures preserved 1:1 in respective SKILL.md files. - All 8 setup sections preserved verbatim with per-question structure (S{n}.Q{m}) + "why I'm asking" rationale. - All 10 triage steps preserved verbatim. - DRAFTS-ONLY rule preserved + amplified (stated in SKILL.md, agent, command, AND enforced by draft_safety_validator.py). - Skip-logic preserved (S4 conditional on S1 surfacing opportunities). - 7-file KB contract referenced verbatim in both directions. REPO STRUCTURE — MULTI-SKILL LAYOUT (CLAUDE.md plugin-schema rule) productivity/email/ ├── .claude-plugin/plugin.json ← skills: ["./skills/inbox-setup", "./skills/inbox-triage"] ├── README.md ← pair overview + 7-file contract diagram ├── agents/ │ ├── cs-inbox-setup.md ← interview persona │ └── cs-inbox-triage.md ← recurring-run persona, DRAFTS-ONLY enforcer ├── commands/ │ ├── cs-inbox-setup.md │ └── cs-inbox-triage.md └── skills/ ├── inbox-setup/ │ ├── SKILL.md ← 8 sections, 25-31 Q discipline │ ├── references/ │ │ ├── kb_file_contract.md ← write-side spec │ │ ├── grill_me_section_walk.md ← discipline + skip-logic │ │ └── voice_calibration.md ← sample-extraction theory + 7 sources │ └── scripts/ │ ├── kb_validator.py ← stdlib: 7-file contract check │ ├── section_progress_tracker.py ← stdlib: 8-section walk state │ └── voice_sample_analyzer.py ← stdlib: pattern extraction └── inbox-triage/ ├── SKILL.md ← 10 steps + DRAFTS-ONLY rule ├── references/ │ ├── kb_file_contract.md ← read-side spec (mirror) │ ├── triage_decision_framework.md ← TAKE/WORTH/PASS/FLAG + 7 sources │ └── drafts_only_safety.md ← NEVER-SEND canon + 7 sources └── scripts/ ├── kb_reader.py ← stdlib: parsed KB load + fail-fast ├── search_window_calculator.py ← stdlib: cadence → window └── draft_safety_validator.py ← stdlib: post-run NEVER-SEND check 20 files, 3,710 lines. Roughly 2x a single-skill slice (capture: 1,560, pulse: 1,643), appropriate for two coupled skills. VERIFIED CLEAN Smoke tests on all 6 scripts: - kb_validator.py: 15/15 PASS on sample (4 core files + h1s + sections + conditional file expectations + triage-log/ dir) - section_progress_tracker.py: full lifecycle (start → record_q → record_ section_done → record_skip → status). Active section advances correctly past S4 skip. - voice_sample_analyzer.py: 5 samples → register/length/hedging/I-vs-We verdicts + opening + sign-off pattern extraction + email-patterns.md output block generation. - kb_reader.py: reads 5/6 sample files (rate-card.md correctly absent), PASS verdict, structured parsing. - search_window_calculator.py: 2x-daily + 14:00 → 9h lookback, window_start 05:00, run_label "Afternoon". Provides Gmail/Outlook/IMAP query templates. - draft_safety_validator.py: PASS on clean log; FAIL on log with `gmail.users.messages.send` (caught by 2 patterns — defense in depth). Action-required guidance fires. CROSS-SKILL CONTRACT ALIGNMENT (PR #657 audit verbatim alignment preserved) Each of the 7 KB filenames referenced multiple times on both sides: email-taxonomy.md: setup=5 triage=9 email-patterns.md: setup=4 triage=8 evaluation-framework: setup=5 triage=7 rate-card.md: setup=5 triage=4 blocklist.md: setup=4 triage=6 tracker.md: setup=4 triage=7 triage-log: setup=6 triage=7 VERTICAL-SLICE STATUS ✓ Slice 1: capture (light prompt-flow, PR #659 merged) ✓ Slice 2: pulse (research-pack, PR #660 merged) ✓ Slice 3: email-pair (workflow-pair — this PR; introduces productivity/) ☐ Slice 4: generator (04-landing) — validates Next.js code template emission ☐ Slice 5: orchestrator/router (13-research) — must reconcile with existing engineering/autoresearch-agent/ After Slice 4 validates the generator shape, only the orchestrator shape remains to be validated. The 6 remaining research-pack skills (litreview, grants, syllabus, patent, dossier, notebooklm) and 02-reflect can then be batched in a single PR each. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json not updated (separate concern; done after all 13 ship) - .codex/skills/inbox-setup + .codex/skills/inbox-triage symlinks not added (auto-sync workflow handles on merge) - engineering/capture/ NOT moved to productivity/capture/ (would break anyone who installed from current path; address in separate cleanup PR) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
alirezarezvani
|
bf5d4c230f | chore: sync codex skills symlinks [automated] | ||
|
Alireza Rezvani
|
31c35d5204
|
Merge pull request #660 from alirezarezvani/claude/build-skills-pulse-research-slice
feat(engineering): pulse skill — Path-B research-pack slice from megaprompt 01 |
||
|
Claude
|
8132c3483a
|
feat(engineering): pulse skill — Path-B research-pack slice from megaprompt 01
Slice 2 of 13: research-pack shape anchor. Validates that the Path-B conversion pattern transfers cleanly to the 7-skill research pack (pulse, litreview, grants, syllabus, patent, dossier, notebooklm) plus the research orchestrator. PR #657's cross-skill consistency audit locked down the Agent Integrity Rules block this slice carries. SOURCE SPEC megaprompts/01-pulse-megaprompt.md (PR #657). The megaprompt is the canonical spec; this plugin is the working implementation. WHAT THE SKILL DOES Multi-source recency research. Takes the pulse of any topic across Reddit, Hacker News, the open web, and (optionally) X/Twitter within a configurable recent window (default 30 days). Forcing 2-4 question grill-me intake clarifies topic specificity, angle (trend / sentiment / problems / opportunities / comparison), time window, and platform scope. Phases 1-3 run in parallel; sequential within each platform; 1 q/sec rate limit per platform. Returns a synthesized briefing with citations, engagement metrics, and cross-platform pattern analysis. RESEARCH-PACK CONVENTION (preserved verbatim per PR #657 audit) - "Agent Integrity Rules" header (3 occurrences) - 1 q/sec per platform (6 occurrences) - three-count tracking sent/received/cited (3 occurrences) - retry once after 3s (2) - 3 consecutive failures → stop (2) - source discipline (1, + repeated by other phrasings) - parallel execution Phases 1-3 (6 occurrences) - trigger phrases match 13-research SIGNALS map: "pulse on" (2), "take the pulse" (2), "current conversation" (3) PATH-B CONVERSION DISCIPLINE - Frontmatter description preserved verbatim from megaprompt spec. - Workflow structure (megaprompt lines 28-44) became SKILL.md section ordering 1:1. - 4 forcing-intake questions preserved verbatim with "why I'm asking" rationale. - All 5 Agent Integrity Rules preserved verbatim. - Error handling table preserved (7 failure modes). - Output format spec preserved with audit-block addition. - SKILL.md ~2,100 words within megaprompt's 1,800-2,500 budget. REPO STRUCTURE (mirrors capture / grill-with-docs 1:1) engineering/pulse/ ├── .claude-plugin/plugin.json ← source.spec field points at megaprompt ├── README.md ├── agents/cs-pulse.md ← persona, three-count enforcer ├── commands/cs-pulse.md ← /cs:pulse <topic> └── skills/pulse/ ├── SKILL.md ← Path-B converted from megaprompt ├── references/ │ ├── research_pack_conventions.md ← 7 sources (Google SRE, Reddit/HN │ API docs, exponential-backoff, │ citation discipline literature) │ ├── cross_platform_synthesis.md ← 7 sources (Brandwatch, Sprout │ Social, Pew, platform-bias studies) │ └── parallel_execution_discipline.md ← 7 sources (Google SRE, RFC 6585, │ backoff theory, Reddit/Algolia │ docs, Brooker on retries) └── scripts/ ├── time_window_calculator.py ← stdlib: window → HN ts + Reddit t= ├── citation_tracker.py ← stdlib: JSON-backed three-count log └── topic_slug_generator.py ← stdlib: slug + duplicate detection 11 files, 1,643 lines. Comparable to capture (1,560) + grill-with-docs (1,747). Heavier than capture by ~80 lines due to denser Agent Integrity Rules + research-pack convention text in SKILL.md. VERIFIED CLEAN - time_window_calculator.py: 30d → Reddit t=month + HN ts=1776211200 + Web after:2026-04-15; 7d → Reddit t=week. Generates exact query templates the skill needs. - citation_tracker.py: full lifecycle (start → record_sent ×2 → record_received ×2 for 20 sources → record_cited ×2 → status → close) works. Audit block output matches the format spec. - topic_slug_generator.py: "Self-Hosted LLM Deployment for Small Teams" → kebab slug, builds output path, correctly detects duplicate and suggests -v2 suffix. - All 3 with --output json: valid JSON. - plugin.json validates; conforms to repo schema with source attribution block. VERTICAL-SLICE STATUS ✓ Slice 1: capture (light prompt-flow, PR #659 merged) ✓ Slice 2: pulse (research-pack — this PR) ☐ Slice 3: workflow-pair (06+07 email) — next, validates the shared-file-contract pattern between coupled skills ☐ Slice 4: generator (04-landing) — validates Next.js code template emission ☐ Slice 5: orchestrator/router (13-research) — must reconcile with existing engineering/autoresearch-agent/ After Slice 3 validates the workflow-pair pattern, the 6 remaining research-pack skills (litreview, grants, syllabus, patent, dossier, notebooklm) can be batched in a single PR — they all share the shape this slice validates. NOT DONE IN THIS PR (intentional) - .claude-plugin/marketplace.json not updated (separate concern; done after all 13 ship) - .codex/skills/pulse symlink not added (auto-sync workflow handles this on merge per existing pattern) https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
alirezarezvani
|
9a47d85f97 | chore: sync codex skills symlinks [automated] | ||
|
Alireza Rezvani
|
352a8825c3
|
Merge pull request #659 from alirezarezvani/claude/build-skills-capture-vertical-slice
feat(engineering): capture skill — Path-B vertical slice from megaprompt 05 |
||
|
Claude
|
48557fa499
|
feat(engineering): capture skill — Path-B vertical slice from megaprompt 05
Vertical-slice install: first of 13 skills derived directly from the
v2 megaprompts (PR #657, merged). Validates the Path-B conversion
pattern (megaprompt → SKILL.md + scaffolding) before batching the
remaining 12 specs.
SOURCE SPEC
megaprompts/05-capture-megaprompt.md (PR #657). The megaprompt is the
canonical spec; this plugin is the working implementation. Drift
between the two is a bug — re-grill with /cs:grill-with-docs if they
diverge.
WHAT THE SKILL DOES
Brain-dump organizer. Catches an unstructured stream of mixed
thoughts/tasks/ideas and transforms it into a 4-section actionable
system (Projects/Ideas, Tasks, Connections, How I Can Help) with zero
information loss. Fast-to-action by design — no upfront intake.
Asks at most ONE mid-organization clarifying question (only when one
item is genuinely ambiguous between task and project). Workspace
detection is real (Glob/Grep) — never fabricates connections.
Compressed output for small dumps (≤5 unrelated items).
PATH-B CONVERSION DISCIPLINE
- Frontmatter description preserved verbatim from megaprompt spec.
- Workflow structure (megaprompt lines 38-48) became SKILL.md
section ordering 1:1.
- 5 operating principles, 4 sections, anti-patterns list, validation
checklist all preserved with minimal restructuring.
- Some megaprompt prose offloaded into the 3 reference files (the
wrapper additions). Net SKILL.md ~1,800 words, within the
megaprompt's 1,400-2,000 word budget.
- Trigger phrases all surfaced verbatim in SKILL.md "Invocation
Triggers" section.
REPO STRUCTURE (mirrors grill-with-docs 1:1)
engineering/capture/
├── .claude-plugin/plugin.json ← source.spec field points at megaprompt
├── README.md
├── agents/cs-capture.md ← persona, no-fabrication enforcer
├── commands/cs-capture.md ← /cs:capture <dump>
└── skills/capture/
├── SKILL.md ← Path-B converted from megaprompt
├── references/
│ ├── workspace_detection.md ← 4 contexts × tactics
│ ├── voice_preservation.md ← 7 anti-pattern examples
│ └── complexity_matching.md ← format-decision table + 3 worked examples
└── scripts/
├── workspace_inventory.py ← stdlib Glob+Grep helper
├── dump_classifier.py ← stdlib heuristic line-classifier
└── complexity_estimator.py ← stdlib full-vs-compressed recommender
11 files, 1,560 lines. Comparable to grill-with-docs (13 files,
1,747 lines) — capture is leaner because it has no separate format
files (Matt's grill-with-docs ships ADR-FORMAT.md + CONTEXT-FORMAT.md
verbatim alongside SKILL.md; capture's spec is fully self-contained).
VERIFIED CLEAN
- All 3 scripts pass `--help`, `--sample`, and `--output json`.
- workspace_inventory.py: correctly Glob+Greps embedded sample tree
(6 files, 5 folders), surfaces auth+login matches with line numbers.
- dump_classifier.py: labels 13-item sample dump (4 context, 4 task,
2 project-component, 2 question, 1 decision). Known limitation:
verbs like "Brief" / "Rewrite" / "Do" not in task-trigger regex
list — heuristic, documented in script docstring.
- complexity_estimator.py: correctly recommends format=full on 14-item
4-cluster dump and format=compressed on 5-item 0-cluster dump.
- plugin.json validates as JSON; conforms to repo's plugin schema
(name, description, version, author, homepage, repository, license,
skills + optional source attribution block).
VERTICAL-SLICE STATUS
This is Slice 1 of 13. Megaprompt shapes covered:
✓ Light prompt-flow (this slice — 02-reflect transfers cleanly)
☐ Research-pack (01-pulse, 03, 08-12) — Slice 2
☐ Workflow-pair (06+07 email) — Slice 3
☐ Generator (04-landing) — Slice 4
☐ Orchestrator/router (13-research) — Slice 5; reconcile with
existing engineering/autoresearch-agent/
After Slice 2 validates the research-pack conversion pattern (which
includes the cross-skill consistency rules audited in PR #657), the
remaining 11 skills can be batched.
NOT DONE IN THIS PR
- .claude-plugin/marketplace.json not updated (separate concern;
would be done in a marketplace-bundle PR after all 13 ship)
- .codex/skills/capture symlink not added (auto-sync workflow handles
this on merge per the existing pattern — see commit
|
||
|
alirezarezvani
|
6a9abc9609 | chore: sync codex skills symlinks [automated] | ||
|
Alireza Rezvani
|
048745d08f
|
Merge pull request #658 from alirezarezvani/claude/install-grill-with-docs | ||
|
Claude
|
30ea6342e5
|
feat(engineering): install grill-with-docs skill (Matt Pocock derivative, MIT)
Installs Matt Pocock's grill-with-docs skill as the fifth Matt-derived plugin in this repo, following the v2.6.0 hybrid-voice import pattern established by write-a-skill / caveman / grill-me / handoff. Upstream: https://github.com/mattpocock/skills/tree/main/skills/engineering/grill-with-docs License: MIT, © 2026 Matt Pocock. Preserved verbatim per MIT. WHAT THE SKILL DOES Docs-anchored grilling session. Where the existing grill-me skill interrogates a plan in isolation, grill-with-docs interrogates a plan against the project's existing language (CONTEXT.md) and recorded decisions (docs/adr/), updating both inline as terminology and decisions crystallise during the session. Matt's three SKILL.md rules preserved verbatim under MIT: - Interview relentlessly, one question per turn, walking the decision tree depth-first. - When a term is sharpened, update CONTEXT.md right there (don't batch). Use the format in CONTEXT-FORMAT.md. - Offer an ADR only when all three are true: hard to reverse, surprising without context, real trade-off. Use the format in ADR-FORMAT.md. REPO STRUCTURE (mirrors grill-me's 1:1) engineering/grill-with-docs/ ├── .claude-plugin/plugin.json ├── README.md ├── agents/cs-grill-with-docs.md ├── commands/cs-grill-with-docs.md └── skills/grill-with-docs/ ├── SKILL.md ← Matt's voice verbatim ├── ADR-FORMAT.md ← Matt's, verbatim ├── CONTEXT-FORMAT.md ← Matt's, verbatim ├── references/ │ ├── ubiquitous_language.md ← 7 sources │ ├── adr_practice.md ← 7 sources │ └── context_md_as_artifact.md ← 7 sources └── scripts/ ├── context_md_linter.py ← stdlib ├── adr_scanner.py ← stdlib └── glossary_code_consistency.py ← stdlib WRAPPER (additions on top of upstream) 1. context_md_linter.py — validates CONTEXT.md against the CONTEXT-FORMAT.md structure: H1, one-sentence description, Language section with bold terms + `_Avoid_:` aliases, Relationships, Example dialogue, optional Flagged ambiguities. PASS/WARN/FAIL per rule. Smoke-tested: positive case PASS 7/7, negative case (broken file) correctly FAILs with 4 WARNs identifying every missing element. 2. adr_scanner.py — walks docs/adr/, checks NNNN-slug.md filename pattern, surfaces numbering gaps + duplicates, validates H1 + body on each ADR, sanity-checks optional status frontmatter, verifies "superseded by ADR-NNNN" targets exist. Smoke-tested: positive case PASS 12/12 on 3 sequential ADRs; negative case (gap + malformed filename + 2-word body) correctly FAILs and surfaces every issue. 3. glossary_code_consistency.py — extracts bold terms from CONTEXT.md, greps codebase, flags two grilling-question seeds: (a) DEAD GLOSSARY — terms defined but never used in code; (b) CODE-ONLY PROPER NOUNS — frequent capitalized identifiers in code that the glossary doesn't define (filtered against a stop-list of generic programming terms). Tunable threshold via --min-frequency. Smoke-tested: sample correctly flags 'Discount' (dead glossary) and 'Subscription' (code-only, at threshold 2). REFERENCES (each cites 7 authoritative sources) - ubiquitous_language.md — Evans (DDD blue book), Vernon (red book), Khononov (Learning DDD), Wlaschin (DDD Made Functional), Brandolini (EventStorming), Avram & Marinescu (DDD Quickly), Fowler bliki. - adr_practice.md — Nygard (2011 ADR essay), Tyree & Akerman (IEEE Software 2005), Zimmermann Y-statements, MADR template, ThoughtWorks Tech Radar, Joel Parker Henderson adr-tools, Spotify Backstage. - context_md_as_artifact.md — Khononov on language drift, Kernighan on naming, Fowler BoundedContext bliki, Fowler UbiquitousLanguage bliki, Confluent data contracts, Brandolini EventStorming, Evans on Conformist / Anticorruption Layer (DDD ch 14). AGENT + COMMAND - cs-grill-with-docs (engineering, opus model) — docs-aware grill persona. Pre-flights the 3 linters before the first question, uses their findings as opening question seeds, enforces the inline-edit + ADR-3-criteria-gate rules. - /cs:grill-with-docs <path-to-plan> — slash invocation. Six forcing-question patterns surfaced (glossary conflict, ADR contradiction, undefined term, code-vs-claim, ADR 3-criteria gate, boundary check). DIFFERENTIATION FROM SIBLING SKILLS - vs grill-me: grill-me grills a plan in a vacuum; grill-with-docs grills against CONTEXT.md + docs/adr/ + codebase. Both ship as separate plugins. - vs caveman: different concern (depth-against-docs vs compression). - vs handoff: different mode (interrogate vs continuation). VERIFIED CLEAN - All 3 scripts pass `--help`, `--sample`, and JSON-output round-trip. - All 3 scripts correctly FAIL on deliberately broken inputs. - plugin.json parses as valid JSON, schema matches CLAUDE.md constraints (name, description, version, author, homepage, repository, license, skills, attribution — no extra fields). - MIT attribution present in: SKILL.md frontmatter + body header, ADR-FORMAT.md HTML comment, CONTEXT-FORMAT.md HTML comment, all 3 reference doc citations sections, plugin.json attribution block, README.md Attribution + License sections, agent + command footers. - File-tree mirrors grill-me's layout 1:1. TOTAL FOOTPRINT 13 files, 1,747 lines (3 markdown specs verbatim from Matt + 3 references + 3 stdlib scripts + 4 wrapper files). Comparable to grill-me's 11 files / 1,205 lines, larger by the weight of the 2 format files Matt ships upstream (ADR-FORMAT + CONTEXT-FORMAT, ~135 lines) and the heavier linter logic this skill requires. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Alireza Rezvani
|
8ebfaea1ad
|
Merge pull request #657 from alirezarezvani/claude/skills-library-megaprompts-c1sQy | ||
|
Claude
|
bd4ddb552e
|
docs(megaprompts): v2 self-audit fixes — 6 findings resolved
Cross-skill consistency audit per the PR test plan surfaced six gaps;
all six are fixed here in a single follow-up commit. Three were real
bugs and three were surface inconsistencies that the orchestrator's
cross-skill validation would flag.
REAL BUGS
(1) 13-research SIGNALS map: the most natural pulse router phrases
("pulse on", "take the pulse", "current conversation") were missing
— invocations like "pulse on AI safety" would fall through to general
fallback instead of routing to the pulse specialist. Added all three
to pulse signals. Also added "grants for" + "litreview" stems. Removed
bracketed placeholders ("research [company]", "research [person]")
from dossier signals because the pseudo-code does literal substring
matching and would never match a placeholder; added an inline comment
explaining why placeholders are intentionally excluded (they would
over-trigger on generic "research X" queries that should ask Q3
domain disambiguation instead). Updated the specialist-registry table
row for dossier and the corresponding anti-pattern entry to match the
new SIGNALS reality.
(2) 01-pulse missing the Agent Integrity Rules block. Pulse predates
the research-pack convention; the 0 hits across rate-limit/three-
count/source-discipline/retry markers meant any orchestrator cross-
skill validator would flag it as divergent. Added the full block:
parallel-across-platforms / sequential-within / 1 q/sec per platform
/ source discipline with training-knowledge tagged / three-count
tracking surfaced in synthesis audit / retry-once-after-3s / stop-
after-3-consecutive-failures / plan-tier detection for Reddit + HN
public APIs. Added the matching 7 rows to the validation checklist.
(3) 06-inbox-setup missing explicit stop condition for its ~25–31-
question intake (heaviest in the library). Added stop-condition spec:
hard ceiling 35 questions, Section 4 skip drops total by 6, intake
closes after Section 8 handoff and is never re-opened — re-running
the skill is the way to change preferences later (detects existing
files, asks per-file replace/merge/skip). One-at-a-time rule applies
across section boundaries.
SURFACE INCONSISTENCIES
(4) 09-litreview + 10-syllabus used "Data Integrity Principles" as
the section header for what 08/11/12/13 call "Agent Integrity Rules".
Same content, divergent names. Normalized both to "Agent Integrity
Rules (research-pack convention)" in section index, body header, and
validation checklist row.
(5) Trigger phrase lists in 08-grants, 09-litreview, 10-syllabus used
smart quotes ("...") inherited from original drafts; rest of the
library uses straight quotes ("..."). Normalized all three.
(6) 08-grants frontmatter description listed "grants for [topic]" as
a trigger but the Trigger Phrases bullet list did not include it.
Added as the lead bullet so both surfaces agree. Same fix applied to
09-litreview (added "litreview on [topic]" bullet) and 10-syllabus
(added "syllabus reading list" bullet) for parity with v2 naming.
VERIFIED CLEAN
- 13-research SIGNALS now: 3 pulse-router phrases present, "grants
for" present, no bracketed placeholders in any specialist's literal
list
- 01-pulse: 3 hits on "Agent Integrity Rules", 2 on three-count, 2 on
1 q/sec, 1 on retry once, 1 on consecutive failures, 2 on source
discipline
- 06-inbox-setup: 1 hit on stop condition (was 0)
- 09 + 10: 3 hits each on "Agent Integrity Rules", 0 hits on "Data
Integrity Principles"
- 08 + 09 + 10 trigger lists: 0 smart-quote occurrences
NOT FIXED (intentional)
- Several litreview triggers ("writing a paper on X", "help me
research X") don't match the 13-research SIGNALS map. This is by
design — those phrases fall through to Q3 domain disambiguation,
where the user picks academic-literature explicitly. Auto-routing
generic "research X" queries would over-trigger; the explicit
fallback path is the right hybrid behavior.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
1e638d3082
|
docs(megaprompts): update orchestrator + README for v2 (13 skills, productivity+research domains)
Updates 00-master-orchestrator.md and megaprompts/README.md to reflect
the completed v2 expansion (13 skills across productivity + research
domains) and the grill-me intake discipline retrofitted across all
megaprompts.
00-master-orchestrator.md:
- Skill inventory split into productivity pack (6) and research
pack (7); old core/email/research grouping retired
- Pack flags renamed: --pack=productivity, --pack=research,
--pack=email-pair (replaces --pack=core, --pack=email)
- Dependency table updated for new skills: web_fetch coverage for
Google Patents/Espacenet/USPTO (11) and multi-source (12);
bash_tool coverage for Lens.org BYOK (11) + SEC EDGAR (12)
- Phase 3 generation note: research (13) must validate AFTER its
specialist registry (01, 08, 09, 10, 11, 12) is generated
- Phase 4 per-skill validation now requires grill-me discipline
(one-at-a-time, forcing format, why-I'm-asking, dependency-
ordered, max-question stop)
- Phase 5 cross-skill validation adds research-13 classification
correctness check (routing signals against specialist triggers)
- New "Grill-Me Discipline" quality-standards section codifies the
Matt Pocock six-rule discipline
- Anti-patterns add the batching + vague-acceptance failure modes
- Failure-mode table adds the grill-me-missing case
- Final deliverable file tree shows the domain split with explicit
productivity / research section comments
megaprompts/README.md:
- File table reorganized: orchestrator / productivity pack / research
pack — separate tables per domain
- Quality standards table gains grill-me intake discipline as
standard #2 (with all 6 sub-principles inlined)
- Pack notes section reorganized to match the new domain split
- New "Email Pair (06 + 07) — Knowledge Base Contract" subsection
documents the verbatim KB file contract at \${WORKSPACE}/Email/
- Portability matrix expanded to 13 skills
- New "Naming Conventions (v2 renames)" table documents every v1→v2
rename with rationale (9 renames; notebooklm preserved)
- Cross-skill validation list adds: no trigger-phrase collisions
with research-13 routing, grill-me presence in every intake,
file-contract alignment for the email pair
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
7b9725735e
|
docs(megaprompts): rename 10 recommended-reading-list → syllabus + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. The new name pivots from the output ("reading list") to the
input ("syllabus") — the input is the more memorable handle since
that's what the user has on their desk.
Adds Phase 0 grill-me intake (3 forcing questions) before parsing the
syllabus, and restructures the existing group-and-confirm step as a
grill-me forcing-options checkpoint:
Phase 0 Q1 syllabus input format forcing choice (file path / pasted
content / image) — each format routes to a different reader
Phase 0 Q2 course audience forcing choice across 6 options (undergrad
intro / undergrad advanced / grad masters / grad doctoral /
professional / mixed) — drives summary jargon level and
discussion-question complexity
Phase 0 Q3 year range forcing choice (1 / 2 default / 5 years) —
drives year_min on every Consensus search
Group-and-confirm checkpoint becomes forcing options: proceed / merge
sections / split section / add section / remove section. Refuses to
start Phase 3 (the Consensus search budget) without explicit user
confirmation.
Bundled script path updates from recommended-reading-list/scripts/ to
syllabus/scripts/. Trigger phrases lead with "syllabus reading list"
while retaining all prior phrases.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
13c74ffded
|
docs(megaprompts): rename 09 literature-review-helper → litreview + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. Drops "-helper" suffix per the v2 rename principle (all skills
implicitly help).
Adds Phase 0 grill-me intake (3 forcing questions) before the
reconnaissance search, and restructures the existing post-Phase-2
checkpoint as a second grill-me moment with forcing options instead
of free-text:
Phase 0 Q1 research-question specificity — refuses vague answers
Phase 0 Q2 framework hint forcing choice (PICO / SPIDER /
Decomposition / hybrid / "you pick") — PICO default, skill
surfaces its own recommendation after recon search
Phase 0 Q3 tentative depth (5/10/20) — re-confirmed at the post-
Phase-2 checkpoint when user has seen the framework breakdown
The post-Phase-2 checkpoint adds forcing sub-area adjustment options
("proceed", "add sub-area on X", "remove and replace Y with Z",
"restart with different framework") and refuses to start Phase 3
without an explicit choice. Plan-tier ceiling surfaced so user can
calibrate depth realistically.
Trigger phrases lead with "litreview on [topic]" + "literature review
on [topic]" while retaining all prior phrases. Validation checklist
adds grill-me discipline requirements for both intake moments.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
9f2c917d84
|
docs(megaprompts): rename 08 consensus-grant-finder → grants + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. NIH-only scope and full 5-facet Consensus + RePORTER POST +
NOSI fetch + 9-section DOCX workflow all preserved unchanged.
Adds Phase 1 as a 6-question grill-me intake replacing the previous
"research idea + 3 multi-select questions" pattern. Each question is
forcing, one at a time, dependency-ordered, with explicit "why I'm
asking":
Q1 research idea — refuses vague answers ("AI for healthcare");
5 Consensus facets depend on precision
Q2 career stage — forcing choice across 5 NIH stages (predoc /
postdoc / early / independent / senior); filters mechanism set
Q3 preliminary data status — forcing choice across 4 levels (none /
pilot / strong / validated); drives mechanism budget
Q4 environment — forcing choice across 4 institution types; affects
scope realism and R15 eligibility
Q5 submission posture — new / resubmission / exploring; resubmission
triggers reviewer-response section in DOCX
Q6 known institute targets — accepts "no preference" as common case;
otherwise validates user hypothesis against RePORTER tally
Stop condition: 6 questions max, no re-opening intake after Phase 2A
starts. Trigger phrases gain "grants for [topic]" while retaining all
prior phrases.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
6c6efa5f95
|
docs(megaprompts): rename 07 email-triage → inbox-triage + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. Updates all companion-skill references from email-setup to
inbox-setup throughout the spec.
Inbox-triage is intentionally LIGHT-INTAKE — it runs on a recurring
cadence with preferences pre-baked into the knowledge base from
inbox-setup. The grill-me discipline here asks ONLY the override
questions that matter THIS run:
Q1 (optional) — search window override, asked only when invocation
is outside normal cadence (e.g., on-demand run after long break
wants 24h window; quick check wants 2h)
Q2 (optional) — category skip override, asked only when user invokes
with skip intent ("just opportunities", "skip newsletters")
Stop condition: max 2 questions; default invocations skip both and
run with KB-default preferences. Validation checklist requires the
light-intake discipline to be stated explicitly so future generators
don't add intake questions that break the recurring-execution flow.
Knowledge-base file contract with inbox-setup remains identical:
core required email-taxonomy.md + email-patterns.md, optional
evaluation-framework.md + rate-card.md, evolving blocklist.md +
tracker.md, Email/triage-log/ directory for per-run logs.
Trigger phrases gain "inbox triage" while retaining all prior phrases.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
3677794569
|
docs(megaprompts): rename 06 email-setup → inbox-setup + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. Updates the companion-skill reference from email-triage to
inbox-triage (paired-skill file contract still matches verbatim — KB
files at \${WORKSPACE}/Email/ unchanged: email-taxonomy.md,
email-patterns.md, evaluation-framework.md, rate-card.md, blocklist.md,
tracker.md, triage-log/).
This is the heaviest grill-me retrofit in the productivity pack
because inbox-setup is the most interview-dense skill. Restructures
all 8 sections to per-question (S{n}.Q{m}) format:
Section 1 (big picture): 6 grill-me questions
Section 2 (categories): 3 questions including yes/mostly/no taxonomy
verification on the proposed list
Section 3 (voice): 6 questions plus the critical S3.SAMPLES
real-sent-email collection (highest-quality voice input)
Section 4 (evaluation framework, conditional): 6 questions, skipped
entirely if S1 didn't surface opportunity emails
Section 5 (blocklist): 3 questions
Section 6 (current state): 3 questions
Section 7 (report preferences): 3 questions
Every question carries explicit "why I'm asking". Forcing format on
multi-choice questions. Section boundaries don't relax the one-at-a-
time rule — drops "Conversational pacing" anti-pattern row in favor
of the harder grill-me discipline principle.
Trigger phrases gain "set up my inbox" + "configure inbox triage"
while retaining all prior email-* phrases. Handoff message at end
of Section 8 references inbox-triage by new name.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
6fdb3dc09f
|
docs(megaprompts): rename 05 brain-dump → capture + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. Capture is intentionally fast-to-action — when the user dumps,
the skill organizes immediately. No upfront intake.
Grill-me retrofit takes the form of a single MID-ORGANIZATION clarifier
asked at most once per dump, only when a genuine ambiguity surfaces
between task and project for a single item. Pattern: identify the most
ambiguous item, ask one forcing question, commit and continue. Multiple
clarifying questions break the dump-and-organize flow that makes the
skill useful — so the spec hard-caps at 1.
Skipping the clarifier entirely is the common case when the dump is
unambiguous.
Trigger phrases updated to lead with "capture this" while retaining
all prior phrases ("brain dump", "let me dump some ideas",
"here's everything on my mind", etc.) as additional triggers.
Frontmatter description explicitly states the fast-to-action discipline
and max-1-clarifier rule.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
96530f745d
|
docs(megaprompts): rename 04 landing-page → landing + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + SKILLS_DIR
path. Output filename shortens from <name>-landing.html to <name>.html
since the OUTPUT_DIR (./landing-pages/ default) already encodes the
context.
Adds Phase 0 grill-me intake (4 forcing questions, one at a time):
Q1 product + 1-2 sentence elevator pitch — refuses vague answers
("app for productivity") and pushes for who-it's-for specificity
Q2 audience register forcing choice (technical buyers / business
buyers / consumers / internal) — dictates copy register, jargon
level, social-proof, CTA framing
Q3 brand overrides (HEX colors + fonts) or "default" — accepts
partial overrides; algorithmic derivation when only primary given
Q4 tone forcing choice (professional / playful / authoritative /
minimal) — prevents tonal whiplash across sections
Recommended defaults per audience baked into Q4 guidance. Stop
condition: max 4 questions; no follow-ups during generation.
Trigger phrases updated to lead with "landing for X" while retaining
all prior phrases. Validation checklist adds grill-me requirements +
new SKILLS_DIR/landing path.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
c405ba7e75
|
docs(megaprompts): retrofit 03 notebooklm with grill-me intake
Name unchanged (brand specificity is the value — drops the LM suffix
would collide with generic notebook intent). Retrofits action-routing
intake as grill-me forcing questions, one at a time, dependency-ordered:
Q1 action commitment (read/extract, add source, generate Studio
output, or create new notebook) — forcing choice; refuses to
start without action declared
Q2 notebook identity (name or URL; "create new" branch asks title
instead)
Q3 action-specific parameter — branches per Q1:
- read: question to ask the notebook
- add source: source type forcing choice across 5 options
- Studio: which output type
- create new: initial sources
Q4 Studio custom prompt detail (angle / audience / length) —
mandatory for action 3, skipped otherwise. Carries 3 concrete
example prompts to model the level of specificity needed.
Stop condition: most invocations exit after Q3; only Studio generation
runs all 4 questions. Validation checklist adds grill-me requirements
including the Q4-mandatory-for-Studio rule and Q1-action-refusal.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
6a6dc4f1f9
|
docs(megaprompts): rename 02 take-a-step-back → reflect + grill-me retrofit
Renames the mega prompt file and updates frontmatter name + output
target path. Skill purpose unchanged: pure-reasoning metacognitive
reflection across the 5-dimension framework (macro, gap, reflective
inquiry, bias check, contextual alignment).
Grill-me retrofit is intentionally minimal here — reflect is a low-
intake skill by design. Adds an OPTIONAL Q1 clarifier (goal / approach
/ assumptions / all-of-the-above) asked only when the invocation
context is too thin to reassess from. Normal invocations mid-rich-
conversation skip the question entirely and run the 5-dimension
analysis directly. Stop condition: max 1 question; default = 0.
Updates trigger phrases to lead with "reflect" while retaining all
prior phrases ("take a step back", "step back", "zoom out", etc.)
as additional triggers so muscle-memory invocations still route.
Validation checklist updated to require the optional-question
discipline and the new SKILLS_DIR path.
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
98e6508fb6
|
docs(megaprompts): rename 01 last-30-days → pulse + grill-me retrofit
Renames the mega prompt file and updates frontmatter name +
SKILLS_DIR + RESEARCH_DIR output paths from last-30-days/ to pulse/.
Adds Phase 0 grill-me intake (max 4 forcing questions, one at a time,
dependency-ordered) ahead of the existing parallel-search phases:
Q1 topic specificity — refuses vague answers ("AI", "tech"); pushes
back with examples
Q2 angle forcing choice across 5 options (trend / sentiment /
problems / opportunities / comparison) — dictates source weighting
Q3 time window choice (7/14/30/60/90 days, default 30)
Q4 platform-scope skip (asked only when Q1 + Q2 suggest some
platforms are off-target)
Updates trigger phrases to lead with "pulse on [topic]" and adds
"take the pulse of [topic]" / "trending: [topic]" / "current
conversation about [topic]". Output header changes to "[TOPIC] —
Pulse (Last [N] Days)" with angle declared in the dated subheader.
Adds vague-topic refusal as the first failure-mode row. Validation
checklist gains grill-me requirements (one-at-a-time, why-I'm-asking
per question, Q1 vagueness rejection, Q2 forcing format).
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|
||
|
Claude
|
c1d2bdc211
|
docs(megaprompts): add 13 research autoresearch hybrid router skill
Adds the research skill mega prompt — the default entry point for the research domain. Implements Architecture C (hybrid router + fallback): deterministic classification of the user's question, delegation to a specialist when confidence is high (≥2 signal matches), and a full plan-decompose-search-synthesize-cite fallback workflow when no specialist fits. Specialist registry: pulse (reddit/hn/x/buzz/sentiment), grants (NIH/ R01/RePORTER/NOSI), litreview (literature review/PICO/SPIDER/meta- analysis), syllabus (course outline/reading list), patent (prior art/ FTO/IP landscape), dossier (entity research/due diligence/meeting prep). Each has documented routing signals so classification is predictable and learnable. Deterministic classification (not LLM-reasoned): documented as concrete pseudo-code. Per specialist, count matched signal phrases; route to argmax if ≥2 signals; route to single-match specialist at exactly 1 signal; otherwise ask Q3 domain disambiguation. Routing decision is ALWAYS surfaced before execution so users can override. Grill-me intake: minimal by design (max 4 questions, most invocations exit after 2). Q1 research-question specificity (refuses vague), Q2 output preference (chat brief vs .docx), Q3 domain disambiguation asked only when classification ambiguous, Q4 fallback scope only when Q3 picked "none of the above". Routes fast; doesn't slow delegation. Fallback workflow (8 steps): decompose into 3-5 sub-questions, source- select per sub-question, sequential 1 q/sec search, fetch-and-extract, per-sub-question synthesis with inline citations, cross-cutting patterns, format-honoring output (markdown brief default or DOCX), three-count audit log with reliability tiers per source. Inherits research-pack conventions: source discipline, three-count tracking, retry-once-after-3s, stop-after-3-failures. 10 documented anti-patterns including the keystone: never LLM-reason classification, never silent-delegate, never run fallback when a specialist fits. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
de75a55c37
|
docs(megaprompts): add 12 dossier decision-grade entity research skill
Adds the dossier skill mega prompt — fourth in the research pack and the second new skill in the v2 expansion. Non-generic by design: refuses to be "tell me about Microsoft" and forces the user to state their hypothesis upfront via mandatory Q4 in the grill-me intake. The dossier tests the hypothesis rather than confirms it, allocating at least 30% of search budget to disconfirming queries. Grill-me intake (max 6 questions, one at a time, dependency-ordered): subject identity + disambiguating identifier (Q1), subject type forcing choice (Q2), purpose forcing choice across 8 options (Q3), MANDATORY hypothesis statement with push-back protocol if refused (Q4 — the non-generic anchor), depth choice (Q5), sensitivity exclusions for journalism + personal vetting only (Q6). Subject-type source matrices: person (LinkedIn, Twitter, GitHub, Scholar, news), company (official site, SEC EDGAR free API, Crunchbase free tier, news, GitHub for tech, Glassdoor sentiment, LinkedIn company page), nonprofit (ProPublica Nonprofit Explorer Form 990s + official), government org (.gov + ProPublica). Optional BYOK MCPs (LinkedIn, Crunchbase, Apollo, Pitchbook, SimilarWeb) flagged in audit log. Hypothesis-driven search discipline: every Phase 4 query classified as supporting or disconfirming. ≥30% disconfirming budget mandatory to prevent confirmation bias. DOCX output (9 sections): executive summary with verdict on hypothesis (SUPPORTED / PARTIALLY SUPPORTED / DISPROVEN / INCONCLUSIVE), identity facts, hypothesis test with explicit supporting + disconfirming evidence, 12-month activity timeline, network signals, reputation signals, red flags tiered (primary/secondary/tertiary source reliability), 3–5 finding-tied conversation hooks (not generic), source provenance + audit log. Inherits research-pack conventions: sequential execution, three-count tracking, retry-once-after-3s, stop-after-3- failures, source discipline, sensitivity-exclusion honored from Q6. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
1f40d7ef1d
|
docs(megaprompts): add 11 patent prior-art + landscape skill
Adds the patent skill mega prompt — third in the research pack and the first new skill in the v2 expansion. Non-generic by design: refuses to be "patent help" and commits to one of five sub-use-cases via the grill-me intake before any search runs. Each sub-use-case dictates a distinct search strategy: - Novelty search: narrow + claims-text focused - Freedom-to-operate: broad + active patents only, jurisdiction-filtered - Competitive landscape: breadth + filer tally + CPC trends - Acquisition diligence: assignee-specific + assignment-chain - Litigation prior-art: target-patent-anchored + pre-priority art Grill-me intake (max 6 questions, one at a time, dependency-ordered): invention description (refuses generic answers), sub-use-case commitment (forcing choice), jurisdictions (Q2-dependent), known prior art anchor, risk tolerance, attorney-status disclaimer (only for novelty/FTO). Search sources: Google Patents (workhorse), Espacenet (global), USPTO PPS (US), Lens.org BYOK (citation graph). CPC/IPC classification follow-up query mandatory after initial hits. Family resolution deduplicates same-invention filings across jurisdictions. DOCX output (8 sections): executive summary + verdict, closest prior art with extracted independent claim 1, patent landscape, citation graph, geographic coverage, FTO flags, sub-use-case-specific strategy + design- around suggestions, audit log. Inherits research-pack conventions: 1 q/sec sequential, three-count tracking, retry-once-after-3s, stop- after-3-failures, plan-tier detection, source discipline, attorney- consultation disclaimer mandatory where Q2 has legal consequences. Trademark, copyright, and trade-secret questions explicitly out of scope. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
ac543b0919
|
docs(megaprompts): add 10 recommended reading list skill
Adds the final mega prompt in the research pack. Generates the recommended-reading-list skill: parses a course syllabus (PDF / DOCX / text / pasted / image), extracts topics + learning outcomes (inferring 3-5 outcomes if missing), groups topics into 6-12 sections, runs 1-2 targeted Consensus searches per section with applied-domain weaving (e.g., "enzyme kinetics food processing applications", not just "enzyme kinetics"), selects 1-3 papers per section (15-25 total), and writes plain-language summaries + Bloom-higher-order discussion questions tied to specific learning outcomes. Uses a bundled JavaScript helper at scripts/generate_reading_list.js for DOCX assembly: takes JSON input + output path CLI args, produces a title page, intro with consensus.app link, boxed learning-outcomes section, numbered hyperlinked papers per section heading with summary + discussion-question lines, and a footer. JSON schema documented in the skill. Inherits the same research-pack conventions as 08 and 09: 1 query/sec sequential execution, plan-tier awareness (3/search free, more on Pro), strict source discipline (only cite this session's Consensus results), three-count tracking surfaced in chat audit summary, retry-once-after- 3s, stop-after-3-failures, group-and-confirm before searching, full untruncated ExternalHyperlink URLs, and 7 documented failure modes. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
df57ab25c6
|
docs(megaprompts): add 09 literature review helper skill
Adds the mega prompt that generates the literature-review-helper skill — research pack #2. Produces a "launching pad" orientation document (not a finished review) via one broad Consensus reconnaissance search, PICO-default framework selection with SPIDER / Decomposition / hybrid fallbacks, an interactive checkpoint (framework breakdown table + depth selector), and a configurable search budget: Quick (5) / Standard (10) / Deep (20) — each fully allocated with explicit reasoning across sub-area, review-article, era-gated, and follow-up searches. Adds cross-search intelligence (repeat-hit papers as foundational signal, recurring authors as dominant groups, citations-per-year as seminal-work proxy) and an 8-section DOCX (topic overview, priority reading order, field timeline, sub-area guides with Boolean strings, key research groups, open gaps with why-they-matter, hyperlinked bibliography, audit log). Inherits the research-pack conventions established by skill 08: 1 query/sec sequential Consensus execution, plan-tier detection from first response, three-count tracking (searches / unique papers / cited), retry-once-after-3s, stop-after-3-failures, strict source discipline, and standard docx library patterns (LevelFormat.BULLET, ExternalHyperlink with full untruncated URLs, dual-width tables, post-save validation). https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
630b41642a
|
docs(megaprompts): add 08 Consensus grant finder skill
Adds the mega prompt that generates the NIH grant-finder skill — first of the research pack. Combines a 5-facet Consensus positioning analysis (established / stakes / current approaches / adjacent methods / gaps) with RePORTER POST queries (narrow AND + broad OR) executed via bash_tool + curl, NOSI fetches via web_fetch, and a 9-section editable DOCX output (executive summary, positioning with gap quotes + draft Significance/Innovation, target institutes, grant opportunities with hyperlinked FOAs, funded overlap, study sections, strategic recs + mandatory program officer rec + submission timeline, references, audit log). Locks in the research-pack shared conventions: sequential 1 query/sec Consensus execution, plan-tier detection via "Found N, showing top M", strict source discipline (only cite this session's tool results, label training-knowledge as reference), three-count tracking (sent / shown / cited), retry-once-after-3s policy, stop-after-3-consecutive-failures, dynamic fiscal-year window, scope+career-stage mechanism matching, and embedded mechanism + submission-timeline reference tables. Scoped NIH-only with non-NIH funders flagged out at intake. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
7d9ca65309
|
docs(megaprompts): add 07 email-triage execution skill
Adds the mega prompt that generates the email-triage recurring-execution skill — the second half of the paired email pack. Knowledge-base file contracts match the 06 email-setup output verbatim: core required (email-taxonomy.md, email-patterns.md), optional core (evaluation- framework.md, rate-card.md), evolving read+update (blocklist.md, tracker.md), plus the Email/triage-log/ directory for per-run logs. Specifies the 10 execution steps: 9-hour-overlap search window, two- query email search (primary + starred-unread), taxonomy classification with skip-reads on lowest priority, conditional sender research, four- category recommendations (TAKE/CONSIDER/PASS/FLAG), draft creation honoring patterns.md voice, format-honoring report delivery with HTML inline-CSS for Gmail, KB updates + observed-override learning loop after 5+ runs, internal triage log, and empty-inbox handling that still surfaces overdue tracker items. Hard rule stated prominently in multiple places: DRAFTS ONLY — NEVER SEND. Provider-agnostic adapter pattern (Gmail MCP, Outlook MCP, IMAP), fail-fast on missing KB (direct user to email-setup), privacy boundary (no credentials in KB), trigger phrases, anti-patterns, 7+ failure modes, and a validation checklist. https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw |
||
|
Claude
|
7c3ddf0b1d
|
docs(megaprompts): add 06 email-setup onboarding skill
Adds the mega prompt that generates the email-setup interview skill —
the first half of the paired email pack. Specifies the exact knowledge-
base file contract at \${WORKSPACE}/Email/ (email-taxonomy.md,
email-patterns.md, optional evaluation-framework.md + rate-card.md,
blocklist.md, tracker.md, triage-log/) that the companion email-triage
skill must consume verbatim.
Documents the 8 interview sections (big picture, categories, voice via
3-5 real sent-email samples, conditional evaluation framework, blocklist
seeding, current state, report preferences, handoff), conversational
pacing discipline (no batched questions, explain why each one matters),
modular skip-logic for non-applicable sections (e.g., no rate card if
no pricing), privacy boundary (never persist credentials), and re-run
safety (per-file replace/merge/skip prompt).
https://claude.ai/code/session_01FEUmeuYhmnxVFq7EZM8ZSw
|