Memory Health

Memory Health Check

Runs validation across all memory files and produces a health report.

---

Step 1: Run integrity tests

python3 ~/.claude/memory/tests/test_memory_integrity.py 2>&1

If all tests pass, report "All green" with the test count.

---

Step 2: On failures, show and offer fixes

For each failed test:

1. Explain what is wrong (plain language)
2. Suggest a specific fix
3. If the fix is simple, do it immediately

Common errors:

• Counts mismatch: recount observations and update MEMORY.md
• Missing in table: add project to the MEMORY.md table
• Invalid type: check for typos in observations
• No details for index: add the missing Details section
• Stale problem: check if the problem has been resolved, update accordingly

---

Step 3: Additional checks (beyond automated tests)

Manually check:

1. Freshness: Any projects with "Last session" more than 30 days ago? Warn.
2. Size: MEMORY.md over 40 lines? Observations over 20 entries? Suggest cleanup.
3. Duplicates: Any repeating observations by summary?

---

Step 4: Report

Memory Health Check: {OK / N issues found}

Tests: {passed}/{total}
Dossiers: {count} ({list})
Observations: {total across all projects}
Open problems: {count}
{warnings if any}

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: vasilievyakov
• Source: vasilievyakov/miracle-infrastructure
• License: MIT

Install and usage instructions live in the source repository linked above.