cmdop instructions
cmdop instructions
cmdop instructions --show-contentcmdop instructions shows the project instruction files Cmdop supplies to
agents. It discovers AGENTS.md and CLAUDE.md from the git root down to your
current directory, expands @imports, and renders them broad-first — the
most-local guidance last, so it wins.
--show-content prints the exact rendered result rather than a summary of
which files were found. Run it from the directory you are about to work in;
the answer depends on where you are.
The same rendered layer goes to the built-in agent and to the driven Claude Code and Codex adapters, so this is what any of them will have read.
Instruction text is advisory input from a repository, and Cmdop treats it as untrusted. It does not grant permissions: what an agent may actually do is decided by the permission rules and the sandbox, which are enforced separately.
When to inspect instructions
Run this before debugging an agent answer that seems to follow the wrong project rule. The rendered output shows which files were discovered, which imports expanded, and what the final instruction layer says at the directory where the agent will work.
Do not paste secrets into instruction files. They are prompt material, not a secret store, and may be shown to more than one assistant adapter.
Cmdop