Every SDLC starts by finding out what already exists. The api-engineer skill routes you into api-discovery, which searches Postman for contracts and owners instead of guessing.

1. Ask Claude Code:

   ✦copy

   ›

   Search Postman for APIs that give me information on movies, starting with my own workspace. Tell me who owns each one and what endpoints it has.

   Paste this into your agent

2. Confirm Claude runs postman search (for example postman search collections "movies") and lists your Movies API collection first, owned by you, with its GET /movies/animation endpoint.

3. Expect more results than your own collection: the search also finds public collections from other teams and public API providers. Treat those as candidates to reuse, not as part of your project — you can bring one in later with postman dependency add without changing your workspace binding.

Tests, mocks, and docs all derive from the contract, so you establish it before building anything. The api-documentation skill captures what a caller needs to know — inputs, responses, and the failure cases — in files you can review in Git.

1. Ask Claude Code:

   ✦copy

   ›

   Inspect the `GET https://api.sampleapis.com/movies/animation` request in my Movies API collection. Write a description on the request that covers the response shape and auth requirements, and save one success example and one error example (for example an unknown path returning 404).

   Paste this into your agent

2. Review the diff Claude makes under postman/collections/. Confirm the request file has a description and that two example files exist under .resources/Get animation movies.resources/examples/.

3. Claude calls the live API to get real responses. The Movies API needs no authentication, so expect the description to say so.

A contract is only as good as the caller's ability to use it, and callers increasingly include AI agents. The ai-readiness skill scores how easily an agent can discover, call, and recover from errors with your API, and the collection-schema-v3 skill lets Claude fix the gaps directly in the collection's YAML files.

1. Ask Claude Code:

   ✦copy

   ›

   Score the AI readiness of my Movies API collection. Show me the score, bucket, and confidence, then fix the highest-impact gaps by editing the collection files, and score it again to show what improved.

   Paste this into your agent

2. Confirm Claude runs postman collection ai-readiness against the local collection, reports the score and bucket (Limited, Fair, Good, or Excellent), and re-runs it after the edits.

3. Confirm Claude also runs postman collection lint on the edited collection so the YAML is still valid.

4. Don't expect the headline score to jump. The last step already added a description and examples, and for a small API with no auth the score starts high (around 100) — what changes is confidence, which moves from low to high once the examples let the checker measure response structure and errors. A score that dips a few points because it now measures something real, such as rate limiting, is a more accurate score, not a regression.