What problem does it solve? During plans and conversations you often derive reusable procedures, but most of them are not worth turning into permanent skills. This Skill applies a significance gate and a redundancy check so only genuinely recurring, non-obvious, generalizable know-how becomes a skill, preventing skill proliferation and description collisions that degrade triggering. ## Core Features & Use Cases - Significance gate: Filters candidates on recurrence, non-obviousness, and generality, plus durability and payoff checks, before anything is built. - Redundancy detection: The skills.py similar command lexically ranks existing skills against a candidate description to surface duplicates or extension targets, preferring folding into an existing skill over creating a new one. - Candidate queue: Borderline or mid-task candidates are appended as stubs to a CANDIDATES.md queue via add-candidate for deliberate later review. - Use Case: After approving a plan that encodes a repeatable deployment workflow, run the distiller to check it against existing skills, then either hand a refined spec to the /add-skill authoring workflow or enqueue it for later. ## Quick Start Ask the assistant to evaluate whether the procedure we just worked out should become a skill, and to check it against existing skills first.
