What problem does it solve? When evaluating an unfamiliar codebase for reuse, wrapping, or forking, default analysis produces bloated inventories that nobody remembers. This Skill forces a hard-capped, one-line-per-primitive distillation so you walk away with a mental model you can explain in a meeting. ## Core Features & Use Cases - Primitive extraction: Identifies reusable contracts, state/storage boundaries, transport boundaries, parsers, and auth capabilities while ignoring UI flows and product copy. - Strict output contract: Produces one summary sentence, 3-7 capability groups, a plain a → b dependency graph, and a "not exposed" section listing what the code does not actually share. - Hard caps and naming discipline: Limits output to 10 primitives, one line each, with opinionated naming and alias lists so the mental model has one term per concept. - Use Case: Before deciding whether a legacy repo can become an MCP server or library, run this prompt to get a one-screen briefing of its actual reusable surface. ## Quick Start Ask the AI to read this repository and distill it into at most 10 reusable primitives with a plain dependency graph using the discover-primitives prompt.
