Skip to content

All slash commands for your AI coding agent

Type / in your AI tool to see the commands; these are the ones to start with:

/work "<what you want>"
/implement-ticket PROJ-123
/review:changes
/commit

Slash commands run inside your AI tool’s chat. The agent-config terminal commands are a separate set — see the CLI reference.

Command What it does for you
/work "<what you want>" Takes a request through refine, plan, implement, test and verify, then reports back. Never commits for you.
/implement-ticket <ticket> The same loop for a ticket; refines the ticket first and asks when something is ambiguous.
/review:changes Reviews your local diff with focused reviewers and merges their findings.
/commit Splits your changes into logical Conventional Commits after showing you the split.
/agent-handoff Summarises a session so a fresh one can pick up where it stopped.

A grouped command shows a menu when you type it bare, or runs one action directly as /group:action.

Group Actions Use it to
/feature explore, plan, roadmap, dev, refactor Shape an idea into a plan before code
/roadmap create, process-step, process-phase, process-full, next Plan bigger work and run it step by step
/review changes, routing Review before the PR, suggest reviewers
/fix ci, pr-comments, quality, comments, references Fix a red CI run, answer review comments, clear linter findings
/audit deps, code, agents Security: advisories, code holes, what your bots can do
/council default, pr, design, optimize, analysis, debate A second opinion from other AI models
/memory add, load, propose, promote, mine-session Keep project memory the agent consults
/cost report, profile Token cost of the session, and how much is loaded
/design-system capture, generate, import, sync Give the agent a design system to follow
/analyze premortem, postmortem, decision, incident, near-miss Think through a risk, a decision or an outage
/refine-ticket, /estimate-ticket — Sharpen and size a ticket before planning

Also on the menu: /project:map draws the code map; /pr:create opens a pull request (Commit and open a PR).

The full list, with each command’s description, is the generated catalog. Which commands you see depends on your profile.

On tools without native slash commands, the commands are installed as text your agent reads; ask for them by name (“run the work command on …”).