dylan-dog

What problem does it solve? Code reviews and spec approvals often miss edge cases, failure modes, and design risks because they rely on a single shallow pass. This Skill systematically hunts for bugs and failure modes before they ship, using structured heuristics instead of gut feel. ## Core Features & Use Cases - Premortem Mode: Analyzes a spec or design before implementation by imagining the feature has already failed, then mapping causes to probability, impact, and mitigations written directly into the spec. - Hunting Mode: Reviews code diffs with a two-pass approach — a triage agent classifies sections by risk, then a hunter agent performs deep adversarial analysis only on high-risk sections, saving tokens. - Shared Heuristic Checklist: Applies systematic categories (SFDIPOT, CRUD, boundary analysis, concurrency, security, mutation thinking, dependency legitimacy, and more) to both modes. - Use Case: Before approving a new payment-flow spec, run a premortem to enumerate failure causes; later, when the implementation diff arrives, the review verifies that each promised mitigation actually exists in the code. ## Quick Start Invoke the skill with /investigate on a spec or code diff to run a premortem or adversarial review.