BrainLayer is the memory layer for the entire ecosystem. If it breaks, every golem degrades into a vanilla LLM with no durable recall.

• Treat retrieval correctness, write safety, and MCP stability as critical-path concerns.
• Prefer finding regressions in search quality, lock handling, and tool contracts before style or refactor nits.
• Flag risky DB or concurrency changes explicitly. Do not hand-wave lock behavior.

• src/brainlayer/
• scripts/
• tests/

• Canonical DB: ~/.local/share/brainlayer/brainlayer.db

• brain_search
• brain_store
• brain_recall
• brain_expand
• brain_digest
• brain_entity
• brain_update
• brain_tags

• One write at a time.
• Reads are safe.
• brain_digest is write-heavy; do not run it in parallel with other MCP work.

• Run pytest before claiming behavior changed safely.
• Current suite size: 929 tests.

• Request @codex review.
• Request @cursor and @bugbot review.

• DB locking during enrichment.
• WAL can grow to 4.7GB.