LLMSkillHub
mirroredCopied from FerroxLabs/ijfw under MIT. FerroxLabs has not claimed this page, and is not affiliated with LLM SkillHub.

ferroxlabs / ijfw-memory-audit

Audit and clean project memory files. Trigger: 'memory audit', 'clean memory', 'memory health', /memory-audit

0
SKILL.md
1.5 KB

Execution

  1. Call ijfw_metrics to get counts and last-update timestamps for all tiers.

  2. Scan .ijfw/memory/ for all .md files. For each, note:

    • File name, size (lines), last modified date.
    • Days since last referenced (use file mtime as proxy).
  3. Categorize entries:
ACTIVE  -- referenced within 30 days
STALE   -- not referenced in 30-90 days
ARCHIVE -- not referenced in 90+ days, or flagged as superseded
  1. Report:
MEMORY HEALTH REPORT
  Total entries: <N>  |  Total size: ~<X> lines
  Active: <N>  |  Stale: <N>  |  Archive candidates: <N>

STALE ENTRIES (>30 days unreferenced)
  - <filename>: <one-line summary>  [last seen: <date>]

ARCHIVE CANDIDATES (>90 days or superseded)
  - <filename>: <one-line summary>  [last seen: <date>]

RECOMMENDATION
  Archive <N> entries to .ijfw/memory/archive/. No data is deleted.
  1. Pruning question: for each entry flagged STALE or ARCHIVE, ask "Would removing this rule cause the agent to make a mistake?" If no, archive. If yes, keep and tighten. Memory that doesn't change behavior is bloat that crowds out memory that does.

  2. Ask before acting:

    Archive <N> stale entries? (y/n -- files move to .ijfw/memory/archive/, not deleted)

  3. On confirmation, move flagged files. Store audit result:

    ijfw_memory_store: memory audit on <date> -- archived <N> entries