secure-api-calls
by smarcombes
Coding Agents & IDEsBrief
Call any API without leaking credentials.
Description
What This Skill Does
Keychains proxies API requests from AI agents, replacing placeholder variables like {{OAUTH2_ACCESS_TOKEN}} with real credentials server-side. The agent never handles actual tokens or keys. Users approve credential access once via biometric on keychains.dev, and all subsequent requests to that provider succeed automatically.
Unlike passing real API keys to agents directly, Keychains keeps credentials in the user's encrypted vault and injects them only at proxy time, so a compromised agent cannot exfiltrate secrets.
When to Use It
- Letting an AI agent list GitHub repos without exposing a personal access token
- Sending Slack messages from an agent using an OAuth token the agent never sees
- Reading Gmail inbox from an automated workflow without storing credentials locally
- Making Stripe API calls from an agent without hardcoding secret keys
- Delegating API access to a cloud-hosted agent without sharing credentials directly
Example Workflow
User asks: List my GitHub repositories
- 1
Run
keychains curl https://api.github.com/user/repos -H 'Authorization: Bearer {{OAUTH2_ACCESS_TOKEN}}' - 2
Keychains returns an approval URL on first use — display it to the user
- 3
Run
keychains wait <approval-url>to poll until the user approves via biometric - 4
Replay the original command — Keychains injects the real token and forwards the request
- 5
Return the API response to the user
JSON list of the user's GitHub repositories
Requirements
- keychains.dev account
Installation
clawhub install smarcombes/secure-api-callsRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.