api-reverse-engineering-toolkitlisted
Install: claude install-skill syall/api-reverse-engineering-toolkit
# API Reverse-Engineering Toolkit
Turn a pile of URLs, captured network traffic, or "here's how the frontend calls this" into a verified, provenance-annotated API spec — and then help the user pick the right format, protocol, and code-generator for what they actually need.
The core discipline of this skill: **never write down a shape you haven't actually observed.** Every endpoint, field, and example in the output spec should be traceable to either a live response you captured, or clearly flagged as inferred (and from what source). This is what separates a reverse-engineered spec that's actually useful from one that silently propagates guesses as facts.
## When this applies vs. when it doesn't
Use this skill for undocumented/internal/third-party APIs the user wants to formalize. It is not a substitute for reading an API's own published OpenAPI/Smithy/GraphQL schema if one already exists — check for that first (common locations: `/openapi.json`, `/swagger.json`, `/.well-known/openapi.yaml`, an `/api/docs` page, a GraphQL introspection query against `/graphql`). Reverse-engineering is for when none of that exists or what exists is stale/incomplete.
## The workflow
This has five phases. Don't skip straight to writing the spec — the recon and verification phases are where reverse-engineered specs usually go wrong.
### 1. Enumerate candidate endpoints
Sources, roughly in order of reliability:
- **URLs the user already gave you.** Start here always — it's ground truth abou