QA Project Memory

QA Project Memory

Purpose

Structured log of all QA activities — bugs found, decisions made, tests executed, regressions detected, environments configured, and known issues tracked. Automatically updated after every QA task with deduplication, cross-referencing, and archive rotation.

Key distinction: memory-bank/ describes WHAT this project IS. docs/qa-memory/ describes WHAT QA HAS DONE.

Commands

Language Policy

Memory Files

All files live in docs/qa-memory/:

See references/file-formats.md for entry formats with examples.

Workflow

init — Initialize Project Memory

1. Create docs/qa-memory/ directory with all files from templates
2. Create _archive/ subdirectory with .gitkeep
3. Populate _index.md with header and search instructions
4. Confirm creation with file count

update — Auto-Record (Post-Task Protocol)

This runs automatically after every QA task. See references/auto-update-protocol.md.

1. Classify — determine which file(s) to update
2. Archive check — verify thresholds before writing
3. Deduplicate — search for existing similar entries
4. Write — append entry in established format (English)
5. Index — add/update entry in _index.md
6. Cross-reference — link related entries by ID

search <query> — Search Memory

Priority order:

1. _index.md — fast lookup by ID, date, keyword
2. Active files — full-text search in docs/qa-memory/*.md
3. Archive — full-text search in docs/qa-memory/_archive/*.md

status — Memory Health

Display per-file stats:

• Entry count / threshold
• Last update timestamp
• Archive file count and total archived entries
• Warnings for files approaching threshold

archive — Manual Archive Rotation

Trigger archive rotation for files exceeding thresholds. See references/archive-strategy.md.

• archive — rotate all files
• archive bugs — rotate only bugs.md
• archive --dry-run — preview without changes

summary [period] — Activity Summary

Aggregate entries from test-log.md for the specified period:

• day — today's entries
• week — last 7 days
• sprint — last 14 days (or custom sprint length)

Output: entry count by type, top bugs, coverage changes, decisions made.

Scope

Can do (autonomous):

• Initialize memory structure (init)
• Auto-record after every QA task (update)
• Search across active files and archive
• Archive entries when thresholds exceeded
• Deduplicate entries
• Cross-reference related entries
• Generate summaries

Cannot do (requires confirmation):

• Delete any memory entry (archive only, never delete)
• Change archive thresholds
• Modify another skill's output

Will not do (out of scope):

• Replace memory-bank/ (different abstraction layer)
• Store credentials or secrets
• Modify source code or test files

Quality Checklist

• Memory files exist in docs/qa-memory/ with correct headers
• Every entry has: date, sequential ID, description, source skill
• All content written in English
• No duplicate entries (deduplication applied)
• Cross-references use correct IDs (BUG-xxx → REG-xxx → ADR-xxx)
• _index.md updated on every write
• Archive thresholds respected
• Archived entries preserve full content (no data loss)

Troubleshooting

References

• references/auto-update-protocol.md — 6-step post-task protocol
• references/file-formats.md — entry formats with examples for all 6 files
• references/archive-strategy.md — thresholds, rotation, index format
• references/integration-hooks.md — which skills write what, hook points
• references/maintenance-guide.md — archive commands, health checks, rebuild