Capability Catalog

Your agent asks DorkOS what it can do, and gets a live answer

Agents guess from docs that go stale. One command asks your running DorkOS instead: the actions it accepts by name, and how risky each one is.

Your Codex agent has no DorkOS tools at all. It runs one command, reads back the current list of actions, and gets to work, without you pasting anything out of the docs.

• One command lists the actions DorkOS accepts by name
• Each one says how risky it is up front
• Run any of them by name with `dorkos call`
• Reaches DorkOS from Codex and OpenCode, which get no in-app tools
• Says what it leaves out, so an agent looks in the right place