Agent skills for Antwork
Connect Antwork, add the skills, and ask your assistant to plan next week or audit your accounts.
What they are
Antwork is an MCP server, so any AI can already draft, schedule and publish through it: the server explains those steps itself. The skills add the multi-step work on top: a content calendar planned around what is already scheduled, one piece repurposed for every platform, a campaign arc, ideas grounded in what performed, an analytics report, and a scored audit.
Install
Claude Code
The repo is also a Claude Code plugin marketplace. Installing the plugin connects the Antwork MCP server and loads every skill and agent in one step.
/plugin marketplace add polynode-co/antwork-skills
/plugin install antwork-skills@antworkComplete the OAuth prompt and you are done. Prefer not to use the plugin system? The repo ships an install script that copies the skills into ~/.claude/ instead. A free Antwork account is enough — there is no paid gate on MCP access.
Codex
Codex reads the same plugin marketplace, so the first two commands add the skills and the Antwork MCP server, and the third signs you in.
codex plugin marketplace add polynode-co/antwork-skills
codex plugin add antwork-skills@antwork
codex mcp login antworkAny agent, with the skills CLI
The skills CLI (skills.sh) installs the skills into Claude Code, Codex, Cursor and other agents. It installs skills only, so connect the Antwork MCP server once as well.
npx skills add polynode-co/antwork-skills
# then connect the server, for example:
claude mcp add --transport http antwork https://api.antwork.io/mcp
codex mcp add antwork --url https://api.antwork.io/mcpclaude.ai
claude.ai has no plugins, so the same skills come as one ZIP. In claude.ai, open Settings, find Skills, and upload it.
Download the claude.ai skill (ZIP)
You still connect Antwork as a connector. The ZIP is rebuilt from the GitHub repo on every change, so this link always serves the current version. On claude.ai the audit runs its five dimensions one after another instead of in parallel.
What's in it
6| Skill | What it does |
|---|---|
| antwork | The router. Sends /antwork <command>, or a plain request, to whichever skill fits. |
| antwork-ship | Run /antwork ship inside a repo: it reads your recent commits, merged PRs or a release, picks the few worth posting, and drafts them per account in each account’s voice. Nothing goes out until you approve it. |
| antwork-calendar · antwork-campaign | Batch scheduling across a week or month around what is already booked, and multi-post launch sequences. |
| antwork-voice · antwork-ideas · antwork-repurpose | Learning an account’s voice from its own posts, generating ideas grounded in what actually performed, and turning one long piece into platform-native versions. |
| antwork-analytics · antwork-audit | Performance reporting, and a five-dimension audit that scores engagement, voice consistency, cadence, content quality and platform coverage. |
| 5 audit subagents + 5 templates | In Claude Code the audit runs as parallel read-only agents, each owning one dimension, with markdown templates for the reports they produce. |