What problem does it solve? Deciding which documents a software project actually needs is error-prone: teams either over-produce a fixed SDLC taxonomy or skip load-bearing artifacts like the PRD. This Skill turns a project idea into a proportional, capability-scoped document plan with producers, skills, tools, and dependency edges for every document. ## Core Features & Use Cases - 10-dimension classification: Classifies the project across archetype, domain, regulatory, scale, security, integrations, UI, data/ML, infrastructure, and business clusters, then derives 10 validator-checked prior-art trigger booleans. - Capability identification: Applies a 4-signal algorithm (reference architectures, domain events, domain nouns, jobs-to-be-done) plus three sizing tests to identify 4-10 product capability areas. - Manifest fan-out: Generates shared and per-capability document entries (feature specs, API specs, data models, wireframes, model cards) with author/reviewer role pairs, authoring/reviewing skill pairs, and an acyclic depends_on DAG. - Amend mode: Re-tailors an existing capability-map.yaml and manifest.yaml on a change request using a T1-T5 change taxonomy with a human gate for structural splits and merges. - Use Case: Given an idea.md for a consumer e-commerce web app, produce a three-key JSON payload (capability_map, product_capabilities, manifest) that a caller writes to capability-map.yaml and manifest.yaml to drive downstream document production. ## Quick Start Use project-document-discovery on my idea.md to decide which documents this project needs and return the capability map and document manifest JSON.
