Skip to content

Commands

Commands are slash commands (for example /work W-012) that start an agent with a ready-made prompt. Anything typed after the command name is passed as its arguments.

Command Agent Purpose
/intake Triage Capture a feature request, bug, refactoring, chore, or spike as a work item.
/board Triage Review the board and suggest the next work item.
/work Orchestrator Run one work item through its type-specific flow.
/product ProductOwner Create or refine an app's product brief and backlog.
/new-package Scaffolder Scaffold packages/<name>.
/new-app Scaffolder Scaffold apps/<name> from the project's app template.
/check current agent Run typecheck, lint, and tests for packages changed against HEAD.

Typical flow

graph LR
  I["/intake"] --> B["/board"]
  B --> W["/work"]
  P["/product"] --> I
  N["/new-package, /new-app"] --> W
  W --> C["/check"]

Command file format

Each command is a Markdown file with frontmatter and a prompt body:

---
description: "What the command does."
agent: Orchestrator   # optional; defaults to the current agent
---

The prompt. Use !`shell command` to inject command output.