infer-mvp-recommendation-from-code

What problem does it solve?

It reconstructs what your codebase already treats as v1 by inferring primary and deferred capabilities from scan-index.json, so product-keeper can produce an MVP scope proposal grounded in real shipped behavior instead of guesswork.

Core Features & Use Cases

• MVP reverse-engineering in brownfield codify: infers “what IS the MVP” from release signals, entry points, churn, and README feature enumeration.
• Evidence-based classification: groups capabilities into primary use cases when churn+entry points/readme support exists, and flags deferred capabilities using thin-code signatures such as empty modules, disabled/commented routes, TODO-heavy files, or v0/experimental path patterns.
• Guardrailed artifact generation: validates inputs, optionally halts if an authoritative MVP artifact already exists in product LTM, writes the inferred scope to STM, and emits both a decision manifest and a resolution trace for checkpoint review.

Example use case: during /codify, take scan-index.json for a mature repository and produce scope/mvp-recommendation.md that lists shipped MVP primary use cases and clearly separates deferred capabilities that are scaffolded or not truly released.

Quick Start

Ask the product-keeper workflow to run infer-mvp-recommendation-from-code for issue N using scan-index.json, and have it write scope/mvp-recommendation.md under the issue’s STM evidence directory.