What problem does it solve?
It prevents an AI agent from getting lost after you’ve already decided to start building by routing to the correct implementation lane (build, debug, test/proof, review, or ship) and forcing evidence-aligned first actions instead of broad exploration.
Core Features & Use Cases
- Action-word routing for implementation: Detects the current main action (build/debug/test/review/ship) and selects the right first path to reduce wrong-first-step churn.
- Minimum-read execution: Reads only the required initial child skill, then optionally adds specialists only when there’s a specific evidence gap.
- Evidence-oriented outputs: Produces code changes, regression proof, runtime validation results, review readiness findings, or release/rollback actions aligned to the chosen lane.
- Fallback when definitions are missing: If scope/acceptance criteria/API or data contracts are unclear, it routes back to project-definition instead of improvising.
Quick Start
Provide the implementation intent (e.g., “debug this failure,” “add failing test,” “review merge readiness,” or “prepare release”) and ask the agent to produce the smallest evidence-backed implementation result.