Know where your agent is working.
A native, colored machine-identity footer. Execution-server identity, not a misleading label from the terminal you connected with.
Tested: Linux / OpenCode 2.2.6. Preview first. No backend restart.
Small, inspected recipes for configuring your ecosystem. Give an agent a URL, choose what you need, and keep control of every change.
Read https://agents.scottynordstrom.org/agents.md, then use the OpenCode machine-identity recipe for this machine. Ask me for the label and color before changing anything.
A native, colored machine-identity footer. Execution-server identity, not a misleading label from the terminal you connected with.
Tested: Linux / OpenCode 2.2.6. Preview first. No backend restart.
No sign-in, agent install, telemetry, or remote execution. Read with a browser or any agent that can fetch HTTPS.
Each recipe brings its own boundaries, version checks, verification, and rollback. Only completed recipes belong on the shelf.
People get clear guides. Agents get Markdown and a small JSON catalog. No special integration required.
Review the source, pick the correct user and machine, then preview exactly what will change.
Apply only what you approved. Check the real outcome and retain a bounded way back.