What problem does it solve? Coordinating multiple specialized AI agents on enterprise work often breaks down without enforced phase ordering, sandbox isolation, and human sign-off, leading to unreviewed code reaching production. This Skill manages the full Requirements → Architecture → Implementation → Validation lifecycle with a mandatory human tech-lead approval gate before any deployment. ## Core Features & Use Cases - Phase-based routing: Directs artifacts between Business Analyst, Architect, Developer, and QA agents using a structured JSON routing command schema, with phase-skip attempts rejected and logged. - Human approval gate: Emits approval requests to a configured review channel; only an explicit acknowledgement from the named human tech lead satisfies the gate — reactions, timeouts, and presence checks do not count. - Sandbox and permission boundaries: Enforces read-only access for BA/Architect/QA agents, isolated sandbox execution for the Developer agent, and zero production credentials for all agents. - Use Case: Modernizing a regulated banking origination flow — the BA agent maps KYC/AML requirements, the Architect designs transaction integrity, the Developer implements in a sandbox with ≥80% coverage, QA runs OWASP adversarial tests, and the human tech lead approves deployment in the review channel. ## Quick Start Orchestrate the OpenClaw squad to route this feature request through BA intake, architecture, implementation, and QA with a human approval gate before deployment.
