Business Logic Extractor

Produces a single Markdown doc capturing the product and domain knowledge embedded in a codebase — entities, relationships, business rules, invariants. Same llms.txt format (H2 sections, annotated code, token-budgeted) but focused on what the system does, not how to use its APIs.

Difference from llms-txt-generator: that skill documents a library's public API from its docs; this one documents a codebase's product behavior from its source. Input is code, not docs. Sections are domain concepts, not API topics.

Execution

Operates on the current codebase. Requires filesystem access.

Pre-flight: Confirm codebase root. Look for manifest (package.json, pyproject.toml, Cargo.toml, go.mod, pom.xml, *.csproj). If missing, ask user to confirm root.

PHASE 1 — Domain Surface Scan

Find where business logic lives. Focus on high-signal locations.