What problem does it solve?
It prevents uncontrolled, open-loop “just code it” behavior by running a closed-loop harness that observes repo state, decides the next control operator, binds the correct skill, dispatches execution to sub-agents, gathers structured evidence, and gates whether the repo is allowed to advance.
Core Features & Use Cases
- Closed-loop state control: Maintain long-term repo baselines and short-term worktrack transitions using Observe → Decide → Dispatch → Verify → Judge → Recover → Close semantics.
- Evidence + Gate enforcement: Separate proof of state (Evidence) from permission to progress (Gate) across review, tests, and policy lanes, then aggregate verdicts.
- Safety and governance guardrails: Enforce baseline branch and git-hash/idempotency rules, require approvals when permissions are at risk, and stop on explicit termination conditions.
- Milestone pipeline aggregation: Track and update milestone artifacts and backlog/history with gate-driven progression, including pipeline repair and milestone integrity checks.
Quick Start
Use this harness skill when you need the system to run the repo’s layered control loop end-to-end and report allowed next routes, verdicts, and stop conditions.
