qa-project-memory

name	qa-project-memory
description	QA project memory with auto-update. Structured log of bugs, decisions, tests, regressions, environments. Automatically updated after every QA task. Archive system with searchable index for large projects. Trigger phrases (EN): "initialize memory", "init qa memory", "find known bug", "search memory", "show what was done", "check regressions", "memory status", "archive memory", "log a bug", "add decision", "memory summary", "what bugs do we know", "update memory", "show test log". Trigger phrases (UA): "ініціалізувати пам'ять", "знайти відомий баг", "пошук у пам'яті", "показати що було зроблено", "перевірити регресії", "статус пам'яті", "архівувати пам'ять", "залогувати баг", "додати рішення", "зведення пам'яті", "які баги відомі", "оновити пам'ять", "показати лог тестів", "що ми вирішили про", "які тестові середовища є".
dependencies	{"recommended":["qa-orchestrator","qa-bug-ticket-creator","qa-flaky-detector"]}

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

Command (EN)	Alias (UA)	Action
init	ініціалізація	Create docs/qa-memory/ structure + update CLAUDE.md
update	оновити	Auto-record (called automatically after each task)
search <query>	пошук <запит>	Search across active + archive + index
status	статус	Memory health: entry count, last update, archive size
archive	архівувати	Rotate entries exceeding thresholds to _archive/
archive --dry-run	архівувати --перегляд	Show what would be archived without acting
summary [period]	зведення [період]	Summary for day/week/sprint
log bug	залогувати баг	Manually add a bug entry
add decision	додати рішення	Manually add an ADR entry

Language Policy

Layer	Language	Rule
Trigger phrases	EN + UA	Both languages trigger the skill equally
Command names	EN + UA aliases	search = пошук, status = статус, etc.
All documentation	English only	SKILL.md body, references, protocols
Memory file content	English only	All entries, field labels, descriptions
Agent responses	User's language	Agent detects and responds in EN or UA

Memory Files

All files live in docs/qa-memory/:

File	Purpose	Auto-updated by
bugs.md	Active bugs with solutions	Bug finders, test-healer
decisions.md	QA decisions (ADR format)	Strategy/plan skills
test-log.md	Task execution log	All writer skills, orchestrator
regressions.md	Regression patterns	Flaky-detector, test-healer
environment.md	Test environments, URLs, accounts	Environment-checker
known-issues.md	Known issues and workarounds	Any skill finding workarounds
_index.md	Searchable index across all files + archive	Every write operation
_archive/	Archived entries (auto-rotated by quarter)	Archive rotation

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

Symptom	Likely Cause	Fix
Memory files not found	init not run	Run init or ініціалізація
Duplicate entries appear	Dedup keywords too narrow	Broaden grep pattern in Step 3
Archive not triggering	Threshold not reached	Check status for current counts
Search misses archived entry	Index not updated	Rebuild index from active + archive files
Wrong language in entries	Non-English content written	All memory content must be English; fix entry
Cross-references broken	ID mismatch	Verify IDs exist in referenced file
_index.md out of sync	Manual file edit without index update	Rebuild: scan all files, regenerate index

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