What problem does it solve? AI coding agents have no memory between sessions, so repositories that rely on tribal knowledge cause agents to repeat the same mistakes. This Skill turns the repository into the agent's memory by making every assumption, constraint, and invariant explicit, machine-readable, and backed by enforcement that fails visibly when violated. ## Core Features & Use Cases - Enforced Context Rules: Pair every "must" or "never" rule in AGENTS.md or CLAUDE.md with a linter, CI check, architecture test, or compilation error so rules cannot silently rot. - Module Boundary Enforcement: Use tools like dependency-cruiser or import-linter to make module boundaries structural constraints rather than documentation conventions. - Feedback Loop for Learning: Convert every bug fix or correction into an AGENTS.md update in the same PR, so the rulebase compounds as the team's immune system. - Use Case: When setting up a repository for AI-agent development, audit it with the included checklist: can a fresh agent, reading only the repo, avoid repeating past mistakes? ## Quick Start Ask the agent to audit this repository's AGENTS.md and module boundaries against the context-explicitness checklist and propose enforced rules for any prose-only constraints.
