Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.pome.sh/llms.txt

Use this file to discover all available pages before exploring further.

A Pome skill is a Claude Code (and compatible) agent skill that drops into your ~/.claude/skills/ directory and gives the agent a named workflow it can invoke on demand. You install them once; from then on the agent handles setup and test execution by itself.

​
Install

pome skills install
This installs /pome-setup and /pome-test into ~/.claude/skills/. By default each skill is symlinked to the canonical copy inside the installed pome-sh package, so npm install -g pome-sh@latest automatically pulls the latest skill content. Pass --copy to drop independent copies instead (CI, Windows without symlink permission, or any environment where symlinks aren’t suitable). Other flags: --force overwrites an existing install; --dest <dir> writes into a custom skills directory (testing and advanced users). Claude Code is supported today; more agents are coming. If ~/.claude/ doesn’t exist on your machine, the command prints the Claude Code install link and exits.

​
/pome-setup

Use when you want to point an existing coding agent at pome for the first time, or after you’ve changed which services your agent talks to. What it does
  1. Identifies your agent (Claude Code, Codex, Cursor, …) and the services it integrates with (GitHub, Stripe, …), then confirms with you.
  2. Adds minimal, non-breaking hooks to your agent so pome can capture tool calls only when you test via pome — your normal runs are untouched.
  3. Registers the agent on the dashboard and prints the URL.
  4. Writes a starter TESTS.md in your repo containing a single “Hello world” scenario tailored to the agent.
  5. Asks whether you want to run it now.
Invoke 
set up my agent to test with pome. Use /pome-setup
Output contract
  • TESTS.md exists in the repo root with at least one scenario.
  • The agent is registered and visible on dashboard.pome.sh under your account.
  • A URL printed back into the chat so you can jump straight to the agent page.

​
/pome-test

Use when you want to test an already-registered agent against the scenarios pome thinks are relevant for it. What it does
  1. Reads TESTS.md from the repo.
  2. Adds the matching scenarios from the bundled library — the ones whose twin matches a service your agent talks to.
  3. Runs them, captures the trace, and reports the score inline.
  4. Surfaces the LLM-judge handoff for any failing run — a short, actionable suggestion to fix or improve the agent.
Invoke 
test my agents with pome
Output contract
  • Each scenario produces a run on dashboard.pome.sh.
  • A pass/fail summary lands in the chat.
  • The full trace lives at runs/<scenario>/<run-id>/ for local runs.

​
Where to browse scenarios

The skills pull from the same library you can browse with the CLI: 
pome scenarios github
pome scenarios github --copy
See GitHub scenarios for the full catalog.

​
Next

GitHub scenarios

The five bundled scenarios /pome-test will pick from when your agent talks to GitHub.

Dashboard

Where registered agents, clones, and runs appear after /pome-setup and /pome-test.