rule-testslisted

<!-- @harness-posture: portable — prose repair methodology for existing rule test material; no harness primitives --> <!-- @dual-pair: rule-tests-protocol --> <!-- spec: agents/rule-test-author.md --> # rule-tests > **Authoritative for:** the Claude Code trigger for the rule-tests write/repair protocol — a thin wrapper; the canonical, AI-agnostic protocol (write/repair flow, single-rule isolation, per-lane honesty map, staleness-consent script) lives in `agents/rule-test-author.md`. > **NOT authoritative for:** project goal — see README.md#why-this-exists; the protocol itself and every per-lane / stack-specific fact — see `agents/rule-test-author.md`; rule CREATION — see `agents/rule-researcher.md` / `/rule-research`. This skill is a thin entry point. The full protocol — read the delivered rule artifact → write or repair its firing test material → run your lane's deterministic verification in single-rule isolation → quote the tool verdict verbatim — is `agents/rule-test-author.md`. Follow it directly; this wrapper exists only so the protocol is reachable by a Claude Code trigger without duplicating its logic (single source of truth, per `dual-implementation-discipline.md §7`). It repairs test material for an EXISTING rule; it never creates a rule (that is `/rule-research`) and never edits the rule artifact itself. ## Run-moment reading obligations {#rule-tests-run} Before repairing, read the authoritative surface for the root you are in: - **Consumer root:** the delive