orchestrator-workflow

What problem does it solve?

It solves the reliability problem of coordinating complex multi-step coding work where an AI must plan tasks, dispatch work to specialized workers, and verify outcomes deterministically instead of hallucinating or skipping required changes.

Core Features & Use Cases

• DB-persistent task orchestration: initializes and maintains a project-local SQLite task plan so tasks can be resumed, audited, and recovered across failures.
• YAML task contracts for dispatch: converts a SPEC or refactoring plan into atomic worker contracts with explicit scope, acceptance criteria, dependencies, and verification commands.
• Deterministic verification + schema validation: requires verification_commands and validates the worker’s returned result against a strict YAML schema before marking tasks done.
• Autonomous recovery chain: retries with previous_attempt_errors, escalates to a read-only doctor, and blocks only when circuit-breaker thresholds and retries require human attention.
• Safe contract enforcement rules: enforces reuse_patterns/forbidden_duplicates, touch-only files, and risk-class behavior (e.g., high-risk no auto-retry beyond the first failure).

Quick Start

Use the orchestrator-workflow skill to set up a project-local orchestrator DB and dispatch worker-coder tasks from YAML contracts until all ready tasks are verified and completed.