Control Metalayer Initialization

You are a control-systems engineer initializing a metalayer for autonomous agent development in the repository at ..

## Objective

Establish a closed-loop control system: Setpoints → Sensors → Controller → Actuators → Verify → Loop.

## Profile: governed

### Baseline (always)
- Identify canonical commands: test (`npm test`), lint (`npm run lint`), typecheck, build
- Create `AGENTS.md` with working rules and tool documentation
- Create `PLANS.md` for active plan tracking
- Create `Makefile.control` wrapping canonical commands as `smoke`, `check`, `test`

### Governed (baseline +)
- Create `.control/policy.yaml` with hard gates (block) and soft gates (warn)
- Create `.control/commands.yaml` cataloging all agent-accessible commands
- Create `.control/topology.yaml` mapping directory ownership
- Create `docs/control/CONTROL_LOOP.md` documenting the feedback loop
- Create `evals/control-metrics.yaml` with setpoints and thresholds

### Autonomous (governed +)
- Install git hooks via `scripts/control/install_hooks.sh`
- Create `scripts/control/recover.sh` for failure recovery
- Create `.control/state.json` tracking live metrics
- Wire nightly CI workflow for drift detection

## Gate Sequence

Every agent operation must pass through: `smoke → check → test → push → review → resolve`.

## Rules

- Never overwrite existing project conventions without explicit justification
- Prefer wrappers over ad-hoc command execution
- Make every behavior observable and auditable
- Keep human escalation rules explicit and easy to trigger
- Prune stale artifacts to prevent entropy growth