A passing test suite can miss a changed response field or a new authorization requirement. Give your agent a clear review boundary, then use a relevant skill to guide a systematic comparison of the old and new behavior.
Prepare the task
Have the proposed diff, API contract, affected clients, and test command ready in your agent’s workspace. Explain which compatibility guarantees matter and which behaviors may intentionally change.
Work through it with your agent
Describe the change
Search for a skill using the kind of review you need, such as API compatibility or regression testing. Review the skill’s description and source before selecting its public release.
Trace behavior through the system
Ask the connected agent to compare request validation, authorization, response shapes, and error handling. Require a concrete input and observable result for each suspected regression.
Prove the findings
Have the agent write a focused regression test for a confirmed problem, run it, and propose the smallest correction. Keep speculative improvements separate from required fixes.
Find a starting skill
Review an API change and write tests for edge cases
Task searches use OpenAI for semantic matching. Keep secrets and repository contents out of the search text. Review the suggested skill’s source and instructions before using it.
What to ask for
A review with file references, reproducible inputs, compatibility effects, and test results. Findings should explain who encounters the issue and what changes for them.
Review the result
Check every reported failure against the intended contract. Run the relevant tests yourself before merging, and verify that the agent did not change unrelated behavior to satisfy an assertion.
Continue the workflow
Connect Claude Code or Codex, then copy the pinned agent prompt from a public skill page.
Related workflow: Agent skills for authentication and authorization reviews.