audit

Knowledge Audit

Re-process previously researched topics through the updated pipeline.

When This Skill Triggers

• "audit my research", "what have I researched?", "re-index everything"
• After updating the research pipeline and wanting to backfill old topics
• "reindex", "backfill", "re-run research on old topics"
• /neo-research:audit explicitly

Step 1: List What's Been Researched

Call the MCP tool to see all topics:

rlm_knowledge_audit()

This scans .claude/docs/ for library directories and reports file counts and sizes. Review the list with the user before re-indexing.

Step 2: Re-index (if requested)

If the user wants to re-process topics through the current pipeline:

rlm_knowledge_audit(reindex=True)

To limit to a single topic:

rlm_knowledge_audit(reindex=True, topic="fastapi")

This reads existing .md files from disk and re-ingests them into the knowledge store. No network calls — fast and offline.

Step 3: Re-fetch (CLI only)

For a full re-fetch from source URLs (network required), use the CLI:

HOME_DIR=$(echo ~)
KNOWLEDGE_CLI=$(ls -d $HOME_DIR/.claude/plugins/cache/shihwesley-plugins/neo-research/*/scripts/knowledge-cli.py 2>/dev/null | sort -V | tail -1)
python3 "$KNOWLEDGE_CLI" audit --refetch

Or limit to one topic:

python3 "$KNOWLEDGE_CLI" audit --refetch --topic fastapi

Rules

• Always show the topic list first before re-indexing. Let the user confirm.
• --reindex is fast (local files only). Recommend this after pipeline changes.
• --refetch hits the network. Recommend this when upstream docs have changed.
• Report results: how many topics, files, and frames were processed.