skill-lifecycle
Witnessed lifecycle for the project skill pack — usage-telemetry sidecar, value/idle-driven auto-archive (never delete, restorable), pin-exemption, journaled reversible transitions. Use to record skill usage, review archive verdicts, archive or restore a skill, or pin one exempt. Max action is archive; nothing is ever deleted. Use when this named workflow matches the task.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 48
- Forks
- 16
01 前言 Front Matter
- name
- skill-lifecycle
- metadata
- { "canonical": "../../../.claude/skills/skill-lifecycle/SKILL.md", "generated-by": "fak project-assets sync" }
- description
- Witnessed lifecycle for the project skill pack — usage-telemetry sidecar, value/idle-driven auto-archive (never delete, restorable), pin-exemption, journaled reversible transitions. Use to record skill usage, review archive verdicts, archive or restore a skill, or pin one exempt. Max action is archive; nothing is ever deleted. Use when this named workflow matches the task.
03 SKILL.md 正文
/skill-lifecycle — witnessed skill lifecycle (never delete)
Keeps .claude/skills/ from rotting without ever destroying work: per-skill
usage telemetry, auto-archive of stale low-witnessed-value skills, and a
journaled, reversible archive/restore path. Hermes-curator parity (#2941), with
fak's twist: transitions are driven by witnessed value, not idle time alone.
All state lives beside the skills, inside the claude lane:
| Path | Role |
|---|---|
.claude/skills/.usage.json |
telemetry sidecar (use/view/patch counts, last activity, state, pin, origin) |
.claude/skills/.lifecycle-journal.jsonl |
append-only archive/restore transition log |
.claude/skills/.archive/<name>/ |
where an archived skill lives — moved, never deleted |
.claude/skills/.archive/.snapshots/ |
pre-archive tar.gz backups |
Hard rules
- Never delete. Archive is a move; a collision is a structured refusal.
- Pinned is exempt — sidecar
pinned: trueorpin: truein SKILL.md front-matter blocks both auto- and manual archive. - Bundled/core skills are never auto-archived. Origin comes from the sidecar, else git (tracked SKILL.md = bundled); unwitnessable = bundled.
- Value guards idle. Auto-archive needs idle >
--idle-daysand witnessed value (use_count + patch_count) <--min-value. An emergency-only skill with proven uses stays live.
Commands
python .claude/skills/skill-lifecycle/skill_lifecycle.py record <skill> [--kind use|view|patch]
python .claude/skills/skill-lifecycle/skill_lifecycle.py status
python .claude/skills/skill-lifecycle/skill_lifecycle.py sweep [--idle-days 30] [--min-value 1] [--apply]
python .claude/skills/skill-lifecycle/skill_lifecycle.py archive <skill> --reason "..."
python .claude/skills/skill-lifecycle/skill_lifecycle.py restore <skill>
python .claude/skills/skill-lifecycle/skill_lifecycle.py pin <skill> | unpin <skill>
sweep is dry-run by default and prints one verdict per live skill with the
reason; --apply executes only the archive verdicts (snapshot → move →
sidecar state → journal entry, in that order).
When invoked as a skill
- Run
statusandsweep(dry-run) and show the operator the verdicts. - Only
sweep --applyorarchiveafter the operator confirms;restoreis always safe to run on request. - After any transition, show the tail of
.lifecycle-journal.jsonlso the operator sees the journaled, reversible record.
Witness
python .claude/skills/skill-lifecycle/skill_lifecycle_test.py — 12 tests,
including the acceptance witness: an idle unpinned agent skill is archived
(bytes preserved, snapshot taken, journaled) while a pinned one is untouched,
and restore reverses the move exactly.
04 同仓库更多 Skills
serve-receipt
On-demand research protocol + deterministic receipt for serving a Hugging Face model on a concrete GPU target — fan out research lanes (model facts/files, engine support matrix, serving recipes, KV/state memory math), compute the tiered feasibility verdict with `fak serve-receipt` (never narrated math), and emit a witnessed launch receipt. Use when asked for an "initial config", "serving receipt", or "how do I serve <model> on <GPUs>".
caveman
Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.
quality-report
Score the repo quality and emit a report card.
ctx-overlay
Compact the context window and prune the memory store when a session grows too long.
quality-report
Score the repo quality and emit a report card.
ctx-overlay
Compact the context window and prune the memory store when a session grows too long.