What problem does it solve? Coordinating AI agents, external providers like Google Jules, NotebookLM, and Stitch, and human approvals across a full software lifecycle is error-prone without explicit contracts, state tracking, and independent verification. This Skill acts as a control plane that routes work through governed phases, enforces machine-readable task contracts, and maintains resumable audit trails. ## Core Features & Use Cases - Phase Routing and Lifecycle Management: Routes natural-language requests to the correct lifecycle phase (setup, discovery, specification, implementation, verification, release, operations, retirement) using a deterministic routing script and durable project state. - Execution Contracts and Approval Gates: Validates JSON execution plans with task DAGs, file ownership, risk levels, and HMAC-signed human approval tokens before any dispatch occurs. - Provider-Aware Delivery: Dispatches bounded tasks to Jules, NotebookLM, Stitch, or local agents through registered transport adapters, with independent review records and SHA-256-bound evidence required before tasks are marked verified. - Use Case: A team adopting an existing repository runs the control plane to audit alignment, register provider capabilities, create an approved execution plan, dispatch bounded implementation tasks to Jules, and require fresh-context verification before integration and release. ## Quick Start Ask the agent to use the agentic-engineering skill to audit and adopt this repository, then follow the routed phase guidance it returns.
