What problem does it solve? Agreed-upon designs discussed in conversation often get lost or drift during implementation, and large multi-task features fail when errors in early tasks propagate uncaught. This Skill persists the design as a structured spec file and then implements it through disciplined TDD, verification, and review gates. ## Core Features & Use Cases - Spec Synthesis: Turns the current conversation into a structured specification (problem statement, user stories with verification commands, interfaces, testing decisions, risks) written to .sixth-sense/spec.md, with self-review, premortem risk analysis, and an optional second-opinion review pass. - Light Workflow: For small scopes (3 or fewer stories, ≤2 files each), runs a single continuous pass of TDD, typechecking, code review, and commit. - Heavy Workflow: For larger work, decomposes the spec into dependency-tracked tasks, dispatches parallel subagents with effort-based model routing, enforces per-task adversarial review gates, and tracks progress in a resumable ledger. - Use Case: After discussing a new caching feature with the AI, ask it to write the spec, then implement it — the Skill decomposes the stories, dispatches subagents for independent tasks, reviews each diff, and commits with a full-branch final review. ## Quick Start Turn our discussion about the caching feature into a spec and then implement it with TDD and code review.
