What problem does it solve? Writing feature specifications that are rigorous enough to drive implementation is slow and inconsistent, and reviewing them against a project's architecture rules requires staff-engineer discipline that is hard to sustain. This Skill structures spec drafting and review so every spec ends with a phased, testable implementation plan. ## Core Features & Use Cases - Skeleton-first drafting with an Open Questions gate: Start with a minimal TLDR plus critical unknowns, block on answers before designing, then iterate into a full spec with architecture, data model, API contracts, and edge cases. - Severity-ranked architectural review: Review specs against the repository's own agent instructions and canonical primitives, producing Critical/High/Medium/Low findings with evidence and corrections. - Autonomous mode for unattended pipelines: Under --autonomous, resolve open questions with the most reversible defaults, mark high-stakes ones as needing human confirmation, and feed the phased plan directly into ot-auto-create-pr. - Use Case: A team needs a spec for adding rate limiting to a login endpoint. The Skill drafts a skeleton, asks blocking questions, researches market leaders, then produces a phased implementation plan that an automation skill executes phase by phase. ## Quick Start Ask the agent to write a spec for your feature, for example: write a specification for adding rate limiting to the login endpoint.
