メニュー
Mine transcripts, .agents, bd, and git for skill diffs, bd updates, or rare wiki entries.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/majiayu000/claude-skill-registry --skill curateコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
LLM token logprobs and calibration. Per-decision confidence, ECE, Brier, reliability diagrams, low-confidence triage.
Analyze LLM token logprobs and calibration. Use for per-decision confidence, ECE, Brier scores, reliability diagrams, and low-confidence triage.
回顾最近 N 天的 Claude Code 使用记录——扫描原始会话数据，按主题分组汇总"我都做了什么"，并从个人操作系统视角输出模式、风险与增删建议。当用户说 /recap、"看看我这几天做了什么"、"回顾一下我最近的会话"、"这两天我用 claude 干了啥"、"活动回顾" 时使用。
| name | curate |
| description | Mine transcripts, .agents, bd, and git for skill diffs, bd updates, or rare wiki entries. |
| practices | ["wiki-knowledge-surface","lean-startup"] |
| hexagonal_role | supporting |
| consumes | [] |
| produces | [".agents/research/*.md"] |
| context_rel | [] |
| skill_api_version | 1 |
| user-invocable | true |
| context | {"window":"fork","intent":{"mode":"task"},"sections":{"exclude":["TASK"]},"intel_scope":"full"} |
| metadata | {"tier":"knowledge","dependencies":[]} |
| output_contract | .agents/research/*.md (synthesis), bd notes, skill diffs (rare); never code mutations |
Role: miner. Input = trinity slice (transcripts,
.agents/,bd,git). Output = skill diffs (proposed), bd updates, rare wiki entries. Never mutates code.
Status (2026-05-08): introduced ADDITIVE in Phase 1 (m6v5.D.1 / soc-78s2v). Existing miners (dream, harvest, forge, compile, retro, post-mortem-mining, flywheel, trace, provenance, defrag) stay until Phase 3 shim conversion (m6v5.D.3). Fix-C smoke (
soc-wb2aa) gates Phase 3.
| Mode | Purpose | Replaces (post-Phase 3) |
|---|---|---|
--mode=dream | Overnight bounded INGEST→REDUCE→MEASURE on .agents/ | /dream |
--mode=harvest | Cross-rig promotion + post-mortem mining + flywheel rollup | /harvest, /post-mortem (mining half), /flywheel |
--mode=forge | Per-session transcript mining (SessionEnd cadence) | /forge |
--mode=compile | Mine→Grow→Defrag→Lint corpus pipeline | /compile |
--mode=retro | Single-session learning capture | /retro |
--mode=defrag | Knowledge defragmentation (overnight) | compile-session-defrag.sh hook |
--mode=watch | In-session drift / loop detection (15-min cadence) | research-loop-detector.sh hook |
--mode=provenance | Decision-trace + artifact-provenance walk | /provenance, /trace |
Mode-budget assertion: 8 modes. Adding a 9th requires demoting an existing one OR refusing the addition (per Fix-F § continuous CI gate).
Anti-goals (hard constraints from architecture):
/rpi or any code-mutating flow./curate --mode=harvest # cross-rig promotion sweep
/curate --mode=forge # mine the most recent session's transcript
/curate --mode=dream --duration=8h # overnight bounded run
/curate --mode=compile # rebuild .agents/ corpus
/curate --mode=retro # capture this session's learning
/curate --mode=defrag # knowledge defragmentation
/curate --mode=watch # in-session drift check
/curate --mode=provenance --bead=soc-X # walk decision trace for a bead
Parse --mode. Each mode has its own scope semantics:
| Mode | Reads | Writes | Cadence (typical loop binding) |
|---|---|---|---|
| dream | .agents/ corpus | .agents/overnight/<run-id>/ summary + per-iteration JSON | overnight (1×/24h) |
| harvest | .agents/ across rigs (~/.agents/learnings/) | ~/.agents/learnings/ (promotion), .agents/harvest/latest.json | daily (1×/24h) |
| forge | session transcripts (~/.claude/projects/<session>/*.jsonl) | .agents/learnings/, .agents/patterns/ | per-session (SessionEnd) or 30m loop |
| compile | .agents/ corpus | wiki/INDEX.generated.md, .agents/compile/<date>.md | weekly |
| retro | this session's transcript + recent diffs | .agents/retro/index.jsonl (append) | per-session (manual) |
| defrag | .agents/ corpus | .agents/defrag/<date>.md (cleanup report) | overnight (1×/24h) |
| watch | last 100 lines of current session transcript | .agents/watch/<date>.md (advisory) | in-session 15m loop |
| provenance | bd graph + git log + .agents/ for given anchor | .agents/provenance/<anchor>.md | on-demand |
For --mode=dream: acquire .agents/overnight/run.lock exclusively. If held, exit with "another curator holds the lock" message; do NOT block.
For --mode=harvest: check the dream lock. If held, defer harvest to next loop fire.
For --mode=forge: per-session lock at .agents/forge/session-<id>.lock. Auto-released at end of run.
Other modes: no lock; concurrent invocation is safe.
Each mode delegates to a body section in this skill (see § per-mode bodies below for outline; full content moves to references/<mode>.md at Phase 2 startup).
Output is one of (priority order, per architecture knowledge-flywheel rule):
.agents/skill-diffs/<date>-<skill>.diff. Operator approves before applying. NEVER writes to skills/ directly.bd note, bd remember, or new bd create for surfaced issues. Direct, no approval queue..agents/research/, .agents/learnings/, ~/.agents/learnings/ (rare; only when knowledge is generally reusable).Every mode appends one line to .agents/LOG.md:
2026-MM-DDTHH:MM:SSZ [curate --mode=<mode>] <run-id> — <short-summary> — outputs: <artifact-paths>
Overnight INGEST → REDUCE → MEASURE iterations until halt:
.agents/ for new artifacts since last run.agents/overnight/<run-id>/summary.{json,md} + per-iteration archiveDetailed body remains inline until Phase 2 extraction.
Cross-rig promotion sweep:
.agents/ across all rigs (paths from ~/.agents/rigs.yaml or fleet config)~/.agents/learnings/ global hub.agents/harvest/latest.json + promoted files in global hubDetailed body remains inline until Phase 2 extraction.
Per-session transcript mining:
~/.claude/projects/<project>/<session>/*.jsonl).agents/knowledge/pending/ for curator reviewDetailed body remains inline until Phase 2 extraction.
Corpus pipeline:
.agents/wiki/INDEX.generated.md (rebuilt), .agents/compile/<date>.md (lint report)Detailed body remains inline until Phase 2 extraction.
Single-session learning capture:
.agents/retro/index.jsonlDetailed body remains inline until Phase 2 extraction.
Knowledge defragmentation:
.agents/ (content-hash + semantic-similarity).agents/defrag/<date>.md with proposed retirements (NEVER auto-deletes; operator approves)Detailed body remains inline until Phase 2 extraction.
In-session drift detection:
.agents/watch/<date>.md; surface high-severity to bd noteDetailed body remains inline until Phase 2 extraction.
Decision-trace walk:
.agents/ mentions.agents/provenance/<anchor>.md with chronological traceDetailed body remains inline until Phase 2 extraction.
skills/rpi/SKILL.md — orchestratorskills/validate/SKILL.md — validator role (paired canonical skill).agents/research/2026-05-08-fix-F-mode-flag-budget.md — mode-cull rationale