What problem does it solve? Writing or revising governance Rules and Agent Skills requires consistent quality, fidelity to user intent, and regression safety, which ad-hoc editing cannot guarantee. This Skill enforces a structured Controller-Author-Reviewer workflow with fingerprinted baselines, automated validation, and up to three review rounds before a Candidate can be accepted. ## Core Features & Use Cases - Role-based workflow: Separates Controller (task alignment and briefs), Author (sole writer of the Candidate), Quality/Change/Correctness Reviewers, and Runner (bounded runtime evidence collection), each with its own contract in references/. - Fingerprint-gated acceptance: Captures a Candidate baseline and fingerprint via a Python CLI, freezes write scope and validation commands, and only completes when all three review perspectives return PASS on the same fingerprint. - Review topology selection: Chooses Integrated Review for closed, low-risk tasks or Independent Review for high-risk, self-hosting, or security-sensitive changes, with escalation from integrated to independent review. - Use Case: When revising a repository's dependency lockfile Rule, the Controller freezes the scope, an Author rewrites it against accepted requirements, three reviewers independently check quality, semantic regression, and correctness, and the change only ships after automated validation passes on the final fingerprint. ## Quick Start Use the write-rules-and-skills skill to revise the existing Rule at rules/dependency-lockfile.md so it also covers monorepo package managers.
