Use Actuant while you build — the agent skill

The agent brief tells an agent how to run an audit when you ask. The skill goes further: it makes your agent reach for Actuant on its own — auditing, fixing and launching your deployed app as a normal part of building, so nothing ships missing the things that decide whether it's found, trusted, and secure.

What the skill does

It's a Claude/Codex/Cursor-style skill with trigger conditions in its description, so your agent loads it at the right moments — after a deploy, before a launch, or when you ask “is my app ready / secure / findable”. Then it runs a loop:

1. 1

   Audit

   Calls audit_app on the deployed URL and hands you the live browser link so you can watch it run.

2. 2

   Read

   Reports the score and the failing checks, heaviest first.

3. 3

   Fix

   Opens the additive and metadata fixes as a reviewed pull request; fixes code-level issues in the repo directly.

4. 4

   Verify

   Re-audits after the fix deploys to confirm the score moved.

5. 5

   Launch

   Submits to directories and starts tracking AI-search visibility.

Install it

The skill lives at a stable URL, compiled from the product itself so it never goes stale:

https://www.actuant.dev/api/skills.md

Claude Code — drop it into your skills directory (user-level shown; use .claude/skills/… to scope it to one project):

mkdir -p ~/.claude/skills/actuant
curl -fsSL https://www.actuant.dev/api/skills.md -o ~/.claude/skills/actuant/SKILL.md

Cursor / Codex / other agents — save the same file into that tool's skills or rules directory, or simply tell your agent: *“fetch https://www.actuant.dev/api/skills.md and follow it while we build.”*

Then connect the MCP

The skill walks your agent through connecting the Actuant MCP the first time it needs it. To do it up front instead, see Connect an agent.