Command surface
The OpenProse command names and what they are for.
Command surface
OpenProse commands are first an agent-session command language.
When you type prose run file.md inside an agent session, the agent should load the open-prose skill and execute the program in-session.
When you use the CLI from a shell, the CLI packages that same instruction and sends it into a supported harness. The CLI does not replace the VM.
Commands
| Command | Purpose |
|---|---|
prose run file.md | run a Contract Markdown or ProseScript program |
prose lint file.md | check Contract Markdown structure before a run |
prose preflight file.md | check dependencies and environment without executing |
prose test path | run OpenProse tests |
prose inspect run-id | inspect a completed run |
prose status [--graph] | summarize recent runs, optionally as a graph |
prose install [--update] | install or update OpenProse dependencies |
prose examples [name] | list or run bundled examples |
prose help | show OpenProse help |
prose migrate file.prose | convert ProseScript toward Contract Markdown |
Shell CLI
The shell CLI is useful when you are outside an agent session.
prose run programs/reviewer.md
prose run programs/reviewer.md --harness claude-sdk
prose doctorThe current CLI supports harnesses such as codex-sdk, claude-sdk, codex, claude, and mock. It can check whether the open-prose skill is installed for the selected provider and can run prose doctor to inspect setup.
For exact install commands, flags, harness defaults, and exit behavior, read the canonical CLI README.
Agent rule
If you are an agent and the user gives you prose ... in chat, do not shell out to the CLI by default.
Load the skill and treat the command as an instruction to the current session.
Next: canonical specs.