You can run npx @api-doctor/cli . by hand all day, and it works. But the real payoff comes from installing it as a skill, so your coding agent runs it for you without being asked.
What a skill actually does
Once it's installed, api-doctor slots itself into your agent's tooling. So after the agent writes or edits some API integration code, it can call the skill, read back a report of what's wrong, and fix it, all while you go get coffee.
Installation
npx @api-doctor/cli install
That writes the skill definition into your Claude Code configuration. Restart Claude Code and it's ready to go. One command, done.
How the agent uses it
After it generates code that touches a supported provider (Resend, Supabase, Auth0, and friends), the agent walks through this loop:
- Run the api-doctor scan
- Read the findings report
- Fix each flagged issue
- Re-scan to confirm it's clean
The whole thing takes a few seconds, and it happens before the code is ever committed.
Supported agents
api-doctor installs as a skill for Claude Code specifically. But the underlying CLI works with any agent that can run a shell command, so Cursor, Codex, OpenCode, and the rest can call npx @api-doctor/cli . directly and feed the output back into their context.
What gets checked
The scan sniffs out which SDKs are present in your project and runs the matching rule sets. There's nothing to configure. It figures out what to check so you don't have to think about it.
See the full provider list for supported integrations.