What problem does it solve? Over time a skill library accumulates duplicated passages, contradictory guidance, stale references to changed source files, and bloated bodies that waste context tokens. This Skill performs periodic garbage collection across all skills so the library stays consistent, current, and within budget. ## Core Features & Use Cases - Dedup and conflict resolution: Merges duplicated content by pattern-key into the single best location and resolves contradictions by replacing stale passages. - Staleness detection by provenance: Tracks file:line@hash citations and flags passages for revalidation when the cited source changes. - Token budget enforcement: Keeps each skill body under 500 lines / ~5k tokens, moving long material into references/. - Safe deletion with gating: Runs skill-lint.mjs --all and skill-eval.mjs --all as regression gates, and requires a second-opinion subagent review before any deletion. - Use Case: After a quarter of skill growth, run this Skill on a schedule to merge overlapping guidance, retire obsolete passages, and emit a reviewable diff without breaking existing evaluations. ## Quick Start Run the meta-skill-consolidate garbage collection over all skills and show me the diff for review before promoting any changes.
