convention-extraction

What problem does it solve?

Teams and AI produce inconsistent code and reviews because important conventions live implicitly in people's heads; this skill captures that tacit knowledge and converts it into explicit, versioned artefacts so AI and new team members follow the same expectations.

Core Features & Use Cases

• Structured interview protocol: A five-question extraction flow that surfaces architectural non-negotiables, frequent AI corrections, security instincts, review rejections, and refactoring philosophy.
• Artefact mapping: Converts interview answers into concrete outputs such as HARNESS.md constraints, CLAUDE.md conventions, and reviewer agent notes with clear role, context, priority, and output format.
• Practical guidance and heuristics: Includes sizing recommendations, anti-patterns, and re-extraction cadence to keep conventions current.
• Use Case: Run during new project setup, onboarding, after a production incident, or when AI output quality varies by prompter to produce enforceable rules and reduce review churn.

Quick Start

Run an extraction interview with senior engineers and map answers to HARNESS.md constraints and CLAUDE.md conventions using the guided five-question protocol.