brain:init

The project brain uses a SQLite database to store indexed knowledge from scanned documentation, memory files, and emails. It ensures idempotent scans and deduplicates entities and decisions, so re-running the initialization safely skips existing entries and only processes new or changed files.

You initialize the project brain by running the skill in your project folder, which triggers an automated scan of CLAUDE.md, memory files, and reference documents to populate the SQLite database and update the knowledge manifest incrementally.

Yes, you can onboard a codebase by scanning reference documents and emails. The initialization process captures knowledge incrementally from these sources, allowing an AI assistant to reuse repository knowledge without re-reading everything from scratch.

The best way to maintain a living project memory is to use a persistent brain database that applies idempotent scans to capture new or changed files. This approach handles re-runs safely, deduplicates decisions, and updates the manifest after writes.

Yes, the project brain initialization supports generating downstream knowledge articles. After the SQLite brain database is populated with indexed knowledge from scanned documentation, it can optionally generate knowledge-engine articles from the captured evidence.