Rule Authoring

Use with $writing-for-agents.

1. Read or create .agents/rules/index.md.
2. Keep one stable invariant per scope-grouped rule file. Require current Code Evidence or a linked accepted architecture decision.
3. Index each rule by stable id, scope, path trigger, task trigger, optional tags, and file. Tags aid discovery; the other triggers decide activation.
4. Keep scoped AGENTS.md files to scope, exact skill routing, and one exhaustive Rule Registry pointer.
5. Evaluate every activated rule as pass, fail, or not-applicable with evidence in the current handoff or work log.

Rule files contain: id, authority, one positive requirement, activation triggers, verification, and exceptions.

For an existing repository, follow migration.md. Completion means the index is exhaustive, every pointer resolves, and project-owned rules remain preserved.