memory-audit
Skill8 starsChanged 35 days agoMITView on GitHub
Runs a memory hygiene pass on the project memory/ directory. Reviews all active memory files against the current codebase and archives or supersedes stale entries. Use before major feature work or when memory files feel out of date. Trigger this when someone says: clean up memory, audit memory files, memory is out of date, stale decisions, memory hygiene, are our memory files current, check if decisions are still valid. Do NOT use when searching for a specific memory entry — use /memory-query instead.
15 other projects use a version of this.
Adopted by
Most recently taken up by
- dvy1987/agent-loomtheir own version, first seen Nov 15, 2026
- nightscout/AndroidAPStheir own version, first seen Nov 1, 2026
- shakhruz/mila-coretheir own version, first seen Nov 1, 2026
- AcKeskin/contexturetheir own version, first seen Nov 1, 2026
- qwwiwi/agentos-skills-publictheir own version, first seen Nov 1, 2026
- adamlinscott/claude-skillstheir own version, first seen Oct 30, 2026
Other versions15 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 15
- FastLED/FastLED★ 7.5ksame name
- nightscout/AndroidAPS★ 1.2ksame name
- dvy1987/agent-loom★ 4same name
- mfang0126/openclaw-skills★ 4same name
- qwwiwi/agentos-skills-public★ 4same name
- adamlinscott/claude-skills★ 3same name
- shakhruz/mila-core★ 3same name
- AcKeskin/contexture★ 2same name
- andreijirohdev-labs/gen-ai-labs★ 1same name
- brenbuilds1/skills★ 1same name
5 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository30 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- backlogSkill
- conventionsSkill
- debugSkill
- devops-azureSkill
- devops-githubSkill
- devopsSkill
- hotfixSkill
- implementSkill
- interview-meSkill
- lint-agentsSkill
- memory-querySkill
- obsidian-briefSkill
- obsidian-captureSkill
- obsidian-dailySkill
- obsidian-recapSkill
- obsidian-searchSkill
- obsidianSkill
- onboardSkill
- pack-reviewSkill
- planSkill
- refactorSkill
- repo-mapSkill
- review-prSkill
- scaffoldSkill
- setup-projectSkill
- smellSkill
- spec-intakeSkill
- system-checkSkill
- whatSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 4 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 12 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Run a memory hygiene pass on the project's memory/ directory.
- Very specific: line 10
If memory/ does not exist in the current project, report that and stop.
- Very specific: line 14
Use Glob("memory//*.md") to discover all memory files.
- Distinctive: line 31
Do not delete files — history is preserved by archiving, not deletion.
- Distinctive: line 39
After reviewing active files, look for unrecorded decisions.
- Very specific: line 43
If the user identifies gaps, offer to create new memory entries for them using the templates in docs/MEMORY-WRITING.md.
- Distinctive: line 52
If no changes were needed, say so
- Distinctive: line 56
Run the conformance check over the corpus, including any file this pass just edited:
- Very specific: line 74
Missing the memory/ directory entirely: If memory/ does not exist, report that clearly and stop.
“Common” is measured: kept by at least 6 in 10 of the 12 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Memory Audit
Run a memory hygiene pass on the project's memory/ directory.
If memory/ does not exist in the current project, report that and stop.
1. Review active files for staleness
- Use
Glob("memory/**/*.md")to discover all memory files. - Skip files with
status: supersededorstatus: archived— these are history only. - For each active file, check whether it still accurately describes the codebase as it exists today:
- Does it reference a service, module, or dependency that no longer exists?
- Has the technology it covers been removed or replaced?
- Does a more recent decision supersede it?
For stale files, edit the fenced lowercase frontmatter in place:
- Context no longer applies → set
status: archived - Superseded by a newer decision → set
status: supersededand populatesuperseded-by:with the newer filename - The issue was fixed → set
status: resolvedand add aresolved: YYYY-MM-DDkey
Write lowercase keys inside the
---fence and nothing else.docs/MEMORY-WRITING.mdis the authority on the dialect. This step previously instructed the bold unfenced dialect retired on 2026-09-02, which would have made this hygiene skill the thing that broke the dialect on every future pass. Change only the keys named above; leave the rest of the frontmatter and the whole body alone.- Context no longer applies → set
- Do not delete files — history is preserved by archiving, not deletion.
1a. Check the repo map for drift
If memory/architecture/repo-map.md exists, read its verified-at-commit stamp and run git diff --name-only <that-sha>..HEAD. If mapped directories have changed, added, or been removed since the stamp, do not archive the map — it is meant to be refreshed, not retired. Instead, flag it in the report and recommend running /repo-map refresh. Only archive the map if the project structure it describes has been entirely replaced.
2. Check for gaps in decision coverage
After reviewing active files, look for unrecorded decisions. Ask the user:
"Are there major architectural decisions, known issues, or pattern choices that have been made but aren't recorded in memory? For example: a technology that was evaluated and rejected, a workaround for a platform constraint, or a deliberate deviation from the conventions."
If the user identifies gaps, offer to create new memory entries for them using the templates in docs/MEMORY-WRITING.md.
3. Report
Output a summary:
- Total active files reviewed
- Files archived (with reason)
- Files superseded (with reason and superseded-by)
- Gaps identified and whether new entries were created
- If no changes were needed, say so
4. Lint the dialect
Run the conformance check over the corpus, including any file this pass just edited:
bash scripts/lint-memory.shExit 0 conforms, 1 is a finding, 2 means the checker's own self-test failed and its verdict is
worthless. Fix any finding before reporting the pass complete — a hygiene pass that leaves the
corpus non-conforming has not finished. Findings name the file and the rule; docs/MEMORY-WRITING.md
is the authority on the dialect.
If scripts/lint-memory.sh is not present in the project, report this step as not applicable and
continue. The script is newer than some checkouts. A missing checker is a stated gap — do not
describe it as a pass, and do not hand-roll a substitute check.
Gotchas
- Archiving too aggressively: A memory file that references a pattern no longer in use may still be valid if it explains why that pattern was abandoned. Read for intent before archiving — "we removed X because Y" is still useful context even if X is gone.
- Missing the memory/ directory entirely: If
memory/does not exist, report that clearly and stop. Do not create the directory — use /setup-project to scaffold the full project structure. - Confusing superseded with archived: Superseded means a newer decision replaced this one — the newer file must be cited in
superseded-by:. Archived means the content no longer applies at all (e.g., references a removed module). Resolved means the issue it records was fixed. These are different statuses with different meanings. - Inviting the user to create too many new entries at once: Step 2 asks about gaps, but memory files are only useful when they capture non-obvious decisions. Do not turn this into a documentation sprint. Focus on decisions that would surprise a new contributor.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

