convention-extraction

Capture implicit team conventions by running a structured five-question interview with senior engineers, then mapping answers into versioned artefacts like HARNESS.md constraints and CLAUDE.md conventions for CI or AI agent enforcement.

Document architectural non-negotiables by applying an extraction interview protocol during onboarding to surface rules, then converting them into explicit reviewer agent notes with defined priorities, roles, and output formats.

Generate CLAUDE.md conventions post-incident by interviewing the team to capture security instincts and refactoring philosophy, then mapping extracted constraints into versioned files for agent or CI enforcement.

Yes, you can use the extraction interview protocol to capture frequent review rejections and map them directly into HARNESS.md constraints and reviewer agent notes, ensuring automated checks enforce historically tacit expectations.

Re-extract team conventions after team composition changes, new project setups, or when AI output quality varies by prompter, using the provided re-extraction cadence heuristics to keep constraints current and reduce review churn.

Limitations include relying on accurate interview responses from senior engineers and avoiding anti-patterns during sizing; constraints may become stale without adhering to recommended re-extraction cadences after team changes.