菜单
Deprecated pointer. The instinct ledger is now the retrospective ledger, managed via retrospective-ledger.ts and driven by /session-retrospective. This skill only redirects the legacy trigger phrases ("instinct を抽出", "学びを蓄積", "extract instincts") to the new entry point.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Use ctx to search local coding-agent history before acting. Use when prior agent sessions may contain relevant insights, decisions, attempts, or transcript context.
Designs and executes QA test cases. Use for "テストして", "QA観点で確認", "動作確認して", "検証して", "QAして", "test this", "QA check", "merge前に確認", or after feature implementation / dependency updates. Also use in plan mode when designing QA or verification plans to apply risk-based prioritization and systematic test design techniques.
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", "ブラウザで確認", "Chromeで開いて", "スクリーンショット", verifying UI, or any task requiring programmatic web interaction. Also use in plan mode when planning browser-based verification to apply correct data extraction strategies and avoid common anti-patterns.
Opt-in project-aware deterministic verification gate. Detects the project type (Nix flake, Deno, Node, Python) and runs build / typecheck / lint / test / security scan / diff-review. Outputs a READY / NOT READY verdict. Not invoked by the default /impl final gate (that default is /completion-audit → /subagent-review). Invoke explicitly when deterministic re-execution is genuinely required. Triggers include /verification-loop / /verify / verification gate / pre-PR check / 検証ループ.
Mandatory completion audit gate. Run this skill ONCE as the final gate task before declaring all work complete. Audits whether implementation and verification evidence are sufficient for the plan's stated purpose. Default mode is a self-audit table by the main session; a fresh completion-auditor subagent is dispatched only when escalation conditions fire. Do NOT run per-task — run only at the final gate. Skipping the completion audit is the #1 cause of false completion claims.
Execute the plan generated by /plan, one task at a time. Per-task — implement, verify acceptance criteria, optionally run simplify-review on large diffs, and mark complete. The final gate is /completion-audit → /subagent-review.
| name | instinct-learner |
| description | Deprecated pointer. The instinct ledger is now the retrospective ledger, managed via retrospective-ledger.ts and driven by /session-retrospective. This skill only redirects the legacy trigger phrases ("instinct を抽出", "学びを蓄積", "extract instincts") to the new entry point. |
This skill has been integrated into /session-retrospective v3. The atomic "instinct" concept has been subsumed into the outcome-tracking retrospective ledger, and the CLI has moved.
If you invoked this skill expecting the previous v2 behavior, please use the replacements below.
CLI path change:
~/.claude/skills/instinct-learner/scripts/instincts.ts~/.claude/skills/instinct-learner/scripts/retrospective-ledger.tsLedger file change:
~/.claude/instincts.jsonl~/.claude/retrospective-ledger.jsonl (migrated via migrate.ts; source renamed to instincts.jsonl.bak)Subcommand map (all backward-compatible):
| Old invocation | New invocation |
|---|---|
instincts.ts add --rule "..." --domain "..." --session "..." | retrospective-ledger.ts add --rule "..." --domain "..." --session "..." |
instincts.ts reinforce <id> | retrospective-ledger.ts reinforce <id> |
instincts.ts list | retrospective-ledger.ts list |
instincts.ts prune | retrospective-ledger.ts prune |
instincts.ts promote | retrospective-ledger.ts promote |
instincts.ts mark-promoted <id> <section> | retrospective-ledger.ts mark-promoted <id> <section> |
| (new) | retrospective-ledger.ts record-proposal <id> --layer <L> --target <path> --plan '<JSON>' |
| (new) | retrospective-ledger.ts verify --transcript <path> [--dry-run] |
All existing subcommands behave identically. The new subcommands (record-proposal, verify) are what make the ledger closed-loop — Phase 1 of /session-retrospective v3 uses them to check whether past proposals actually changed behavior.
When asked to "extract instincts" or "instinct を抽出" or "学びを蓄積":
If the user wants the full self-improvement flow (extract from this session, propose changes, track outcomes) → run /session-retrospective instead. The retrospective handles instinct creation as part of Phase 4 Present & Track.
If the user wants to add / list / reinforce a specific ledger entry directly → use the retrospective-ledger.ts CLI (see the subcommand map above).
If the user is asking about the ledger itself (contents, confidence scores, promotion candidates) → retrospective-ledger.ts list / promote.
The v2 instinct-learner modeled each rule as a rule-only entry — no link to the proposals derived from it, no record of whether those proposals prevented future misbehavior. The v3 ledger extends each entry with proposals[] and outcomes[] so that /session-retrospective v3 Phase 1 can auto-verify and feed recurrence back into Phase 2 as escalation signals. The CLI rename reflects that the file is no longer just instincts — it is the retrospective source of truth.
See ~/.claude/skills/session-retrospective/SKILL.md for the full v3 architecture.
/session-retrospective — primary entry point/cross-session-analysis — 100+ session Gemini-based analysis (unrelated to ledger)