| name | cleanup |
| description | Clean up stale discovery research and outdated findings. Run periodically to keep Researches/discovery/ fresh and relevant. |
| user-invocable | false |
Cleanup Skill
Keep discovery research fresh — remove stale data, consolidate findings
When to Run
- When
Researches/discovery/ files are older than 14 days
- When findings reference outdated trends or tools
- When files grow beyond ~200 lines (consolidate, don't just append)
Cleanup Process
1. Check Staleness
Read each file in Researches/discovery/. Check the Last updated: header.
| Age | Action |
|---|
| < 7 days | Keep as-is |
| 7-14 days | Review — remove anything obviously outdated |
| > 14 days | Full refresh — re-research and rewrite |
2. Remove Outdated Entries
Delete findings where:
- The trend/tool no longer exists or is no longer relevant
- The demand signal was temporary (one-time event, not sustained)
- The repo already fully covers the use case (gap is filled)
- Competitor analysis references repos that are now archived/dead
3. Consolidate
If a file exceeds ~200 lines:
- Merge duplicate/overlapping entries
- Remove lowest-priority items (Low priority + low automation score)
- Keep top 20 findings per file, ranked by impact
4. Cross-Reference with Repo
Check if discoveries have been acted on:
- Scan recent
git log for auto-commits that match discovery priorities
- Mark completed items (prompt was created/improved)
- Remove completed items from the priority list
Storage Rules
- Always preserve the
Last updated: header — update it after cleanup
- Never delete the files themselves, only prune contents
- Keep at least 5 entries per file (don't over-prune)
