Install the Skill in other clients

The standard, not the vendor

The SKILL.md format is open. Any client that lists itself at https://agentskills.io/clients can load the Autousers Skill — the file you install is the same one regardless of which agent will be reading it.

The two universal install methods

Almost every agentskills.io-compliant client accepts one of these two install shapes — pick whichever matches your client's docs:

Method A — copy the directory

If the client reads skills from a filesystem directory (the most common pattern, used by Claude Code and many community clients):

# 1. Get the bundled skill from the npm package
npm install --no-save @autousers/mcp
cp -r node_modules/@autousers/mcp/skill/autousers <client-skills-dir>/autousers

# 2. Or use --target with the install-skill subcommand:
npx -y @autousers/mcp install-skill --target=<client-skills-dir>

Replace <client-skills-dir> with your client's path (e.g. ~/.cursor/skills/, ~/.codex/skills/)

Method B — upload the zip

If the client accepts skills as zip uploads (the pattern used by Claude.ai and several SaaS clients):

1. Download autousers-skill.zip from the @autousers/mcp GitHub release page
2. Open your client's settings or skills/extensions panel
3. Upload the zip via the client's UI
4. Restart the session if the client requires it

Connect the MCP

The Skill assumes the @autousers/mcp MCP server is reachable from the same client. Most agentskills.io-compliant clients also speak MCP — see Connect with MCP for client-specific configuration. If your client supports Skills but not MCP, the Skill loads fine but cannot drive Autousers; you would need to provide the Autousers REST API to the agent through some other tool surface.

Verify the install

• The client's skills panel (or equivalent) lists 'autousers' with the description from SKILL.md
• The agent calls autousers_list before evaluations_create and presents persona candidates rather than picking on its own
• When you ask for a new eval, the agent uses dryRun: true on the first call and shows you a cost estimate before committing

Reporting client-specific issues

If the Skill loads but workflows misbehave on a specific client, file an issue at https://autousers.ai/contact with the client name and version, and a short transcript. Behaviour differences usually trace to (a) the client's MCP transport implementation or (b) how the client surfaces SKILL.md to the model — both are upstream of our content.