What problem does it solve? High-risk changes like auth flows, billing logic, or schema migrations often get finalized without durable proof of review, making work hard to audit, resume, or hand off. This Skill enforces a manual-first evidence pack so risky work is reviewable before it is marked done. ## Core Features & Use Cases - Six High-Risk Class Definitions: Classifies work across auth/identity, billing/credits, schema migrations, public API contracts, deploy/runtime behavior, and permission/secret/trust-boundary logic. - Five-Artifact Schema: Defines risk-gate.json, context-snippets.json, verification.json, review-decision.json, and adversarial-validation.json, colocated in the plan's task folder harness directory. - Validation Scripts: Ships two Node.js validators that check artifact presence, required fields, risk levels, and explicit APPROVE/REJECT review decisions. - Use Case: Before merging a change to Stripe credit accounting, generate the evidence pack, run the validator against the harness directory, and record an explicit reviewer decision instead of relying on narrative updates. ## Quick Start Ask the agent to generate a risk evidence pack for the current high-risk change and validate it with the bundled script before finalizing.
