ML.ai CodeSkills & Commands
Slash commands
The built-in slash commands, your own OpenCode commands, and why some entries appear greyed out.
Built-in commands
| Command | What it does |
|---|---|
/review | Reviews code: with no argument, reviews all uncommitted changes (staged, unstaged, and untracked files via git status); given a commit hash, reviews that commit; given a branch name, diffs your current branch against it; given a PR URL or number, pulls it with gh pr view/gh pr diff (including cross-repo) and reviews it. Reads the full files being changed, not just the diff hunks, checks AGENTS.md/CONVENTIONS.md/.editorconfig when present, and traces callers/consumers of a changed API rather than judging a hunk in isolation. Review only: it doesn't apply fixes unless asked separately. |
/skills | Narrows the list to skills |
/background | Opens the background drawer, shows the live run count |
/mcps | Opens MCP settings, shows a count of servers needing attention |
/init | Runs the engine's guided AGENTS.md setup |
/<skill-name> | Selects a specific skill directly |
Your own commands
Slash commands you've configured through OpenCode show up in the same list, when ML.ai can run them safely.
Greyed-out commands
A command that appears but can't be selected always states why. This happens for one reason:
- The command runs shell code from its own template. Running it would execute that code without going through the normal approval gate, so ML.ai refuses rather than making a silent exception to the permission system.
This is about what the engine reports back to ML.ai, not a judgment call about whether the command itself is safe. See limitations for how this is tracked.