Audit and rewrite Claude Code auto-memory files to fit load and recall caps.
npx skills add https://github.com/21StarkCom/stark-skills --skill stark-memory-21starkcom
Please help me install this Agent Skill. Skill: stark-memory Source: https://github.com/21StarkCom/stark-skills/tree/main/skill/stark-memory Command: npx skills add https://github.com/21StarkCom/stark-skills --skill stark-memory-21starkcom
--apply is passed. - Lossless rewriting: Shortens index lines, splits or tightens over-cap topic files only when no facts are lost, and moves cross-repo facts into the correct project's memory dir. - Use Case: Your MEMORY.md has grown past 25,000 characters and new memories stop appearing at session start. Run the audit, review the dry-run plan, then apply it to shorten index lines and split dense notes. ## Quick Start Ask the assistant to audit my Claude Code memory with /stark-memory, review the dry-run report, then rerun with --apply to fix over-cap files.High-intent search queries and answers about installing and using this skill.
Run the memory audit to measure the index against the 200-line and 25,000-character caps, then apply the rewrite phase. It shortens index lines over 150 characters to a title plus one hook while preserving every link target.
Claude Code loads only the first 200 lines or 25,000 characters of MEMORY.md, whichever comes first, and silently drops the rest. The audit report names the first dropped line so you can see exactly which entries are invisible.
No. Dry-run is the default and the measuring tool is read-only. Rewrites happen only when you pass --apply, and all writes go through Write/Edit tools rather than shell moves.
Yes. The report flags cross-repo facts as strong or weak based on foreign repo slugs. With --apply, strong facts are moved into the target repo's own memory directory after operator confirmation.
The skill never blind-tightens dense notes. It either splits the file into separate one-fact files with new index lines, or leaves the original in place and flags it for a manual split.
organvm-i-theoriaConsolidates agent memory files by merging duplicates, fixing stale facts, and pruning the index.
anorthamRestore session context with Goldfish recall, Julie workspace indexing, and Sherpa workflow guidance.
DiologIRRe-run whole-corpus quality reviews of diolog-swe-bench and diff against previous reviews.
ano4lAnalyze Claude Code auto-memory to identify promotion candidates and stale entries.