What problem does it solve? Projects accumulate implicit design knowledge scattered across code, comments, and prose, with no durable record of what the system actually is. This Skill autonomously reads the entire codebase and produces an as-is design corpus — concept, story, and decision catalogs — plus an issue intake of open questions, without overwriting anything humans have already refined. ## Core Features & Use Cases - Two-phase autonomous discovery: Phase 1 writes thorough as-is scaffolding under .ok-planner/design/_discover/; Phase 2 extracts durable concept, story, and decision catalogs, each with a produce → review → fix loop capped at 3 cycles. - Issue intake filing: Ambiguities, muddiness, and the run's own confessed uncertainty become issue files under .ok-planner/issues/ with kind: "discover" and status: open, ready for /verify-issues and sprint resolution. - Safety guarantees: Aborts rather than overwrite non-empty human-edited catalogs, runs end-to-end with no mid-run prompts, and regenerates catalog TOCs (concepts.md, stories.md, decisions.md) for fast lookup. - Use Case: Point it at a legacy project with no design documentation; it produces a complete as-is corpus of load-bearing concepts, user-observable stories, and technical decisions, plus a backlog of judgment questions for the owner. ## Quick Start Run /discover-design in a project that has no .ok-planner/design/concepts directory to bootstrap the full design corpus.
