Sign InGet Started

Slash Commands

Run Hamster delivery commands directly from your AI coding agent — /ship a brief, sync your skills, and drive workflow without alt-tabbing back to the browser.

Overview

Slash commands are short verbs your AI coding agent recognizes when it is working in a Hamster-synced repo. They live in the agent surface (Claude Code's command palette, for example) and call into the Hamster CLI under the hood. Because they are wired up by the CLI itself — not authored in your repo — the list of available commands stays current with every hamster sync.

The most important slash command is /ship. It is how Hamster's cloud delivery agents pick up a brief and drive it to a pull request. Other slash commands cover sync hygiene, skill management, and quick lookups.

How It Works

  1. Sync your repo — Run hamster sync (or hamster sync --watch) at least once. The CLI generates a Claude skill at .claude/skills/hamster-project-context/SKILL.md that registers the available slash commands inside Claude Code.

  2. Open your agent — Start a session in Claude Code, Cursor, Codex, or another agent that reads from .agents/skills/ or .claude/skills/. The agent picks up the Hamster skill and the commands it exposes.

  3. Type the command — In Claude Code, type / to open the command palette, then pick a Hamster command (or type its name). Other agent surfaces use their own command UX, but the underlying commands are the same.

  4. Pass arguments — Most commands take a brief identifier — a display ID like HAM-123, a slug, or a full Studio URL. The CLI parses any of those formats.

Where Slash Commands Run

Surface Command UX
Claude Code / opens the command palette; Hamster commands appear under the hamster-project-context skill
Cursor Available through the agent's command surface when the universal skills directory is connected
Codex / Copilot / Continue / Windsurf Each agent uses its own command surface; the same skill files drive the available commands
Hamster Cloud Agents The cloud sandbox invokes commands like /ship non-interactively as part of delivery runs

The slash commands themselves are generated by hamster sync from the skills in your workspace, so they update automatically when your team adds or changes a method or skill that introduces new commands.

The Most Useful Commands

/ship <brief-id>

Hands a brief to Hamster's cloud delivery system and drives it through to a pull request.

When you run /ship HAM-123, Hamster's sandboxed cloud agent:

  1. Pulls the latest brief, task, and context state from your workspace using the CLI inside the sandbox.
  2. Opens the brief and walks the task list in priority and dependency order.
  3. Drives an AI coding agent inside the sandbox to do the work, with the brief, plan, methods, and blueprints all on disk as context.
  4. Opens a pull request when the work is done — or pauses for clarification when it hits an ambiguous decision.

You stay in your editor while it runs. Streaming progress appears in the brief's delivery thread on Hamster, and the resulting PR links back to the brief automatically.

hamster task status <id> <status>

Move a task to todo, in_progress, or done from the terminal. Useful inside agent runs where the agent should mark its own progress without leaving the shell.

hamster task assign HAM-123
hamster task status HAM-123 in_progress
# … work happens …
hamster task status HAM-123 done

hamster sync and hamster sync --watch

Pull the latest brief and task state into .hamster/. Run it once at the start of a session, or use --watch to keep local files in sync with what your team is doing in the browser.

When the agent cannot find a task or brief that you mention, the right fallback is hamster sync first, then look again — the local files may simply be stale.

hamster skills sync

Bidirectional sync for the shared skills your AI agents read. Pulls new and updated skills from your workspace and offers to push back any skills you (or your agent) edited locally. Asset-aware — changes to bundled references and assets are detected too.

hamster skills sync                  # pull and push
hamster skills sync --pull-only      # pull only
hamster skills sync --push-only      # push only

If a skill changed both locally and remotely since your last sync, you will be prompted to keep local, use remote, or skip. Non-interactive runs auto-skip conflicts so nothing is force-overwritten silently.

hamster skills agents

Pick which IDE or agent directories should be linked to the canonical .agents/skills/ location. Run this once when you onboard a new agent or change tools.

hamster brief create, hamster blueprint create, hamster method create

Create new briefs, blueprints, or methods from the terminal. Useful when an AI agent identifies that a new method or blueprint should exist and wants to draft one without context-switching to the browser.

hamster brief create --title "Onboarding redesign" --content "## Context..."
echo "## From stdin" | hamster brief create --title "Piped Brief" --content -

Always pass --content — the markdown is rendered straight into the Studio editor so your collaborators see it immediately.

hamster brief get and hamster task get

Print the full content of a brief or task in the terminal without switching to the browser. Both accept display IDs (HAM-123) or UUIDs.

Tips

  • Pair hamster sync --watch with your editor session so /ship, task get, and brief get always see the latest state.
  • Slash commands respect your role. A Reviewer cannot ship a brief or push skill changes; the command surfaces this back as a permission error rather than failing silently.
  • Skill changes pushed via hamster skills sync go through Hamster's review and version flow — your team can see the diff before it lands.
  • The Hamster CLI auto-checks for updates and prints a notice when a newer release is available. Run hamster --version if you want to check manually.
  • If a slash command is missing inside Claude Code, run hamster sync once to regenerate the skill file.

Related