Agent integrations

The proxq skill teaches an AI agent to use a proxq instance that is already running: submit a request, poll the job, fetch the content and cancel a job it submitted. It does not deploy or reconfigure proxq.

The skill works in any agent that reads .agents/skills/ (such as peen), and installs natively in the clients below. It reads the instance address from PROXQ_URL.

Claude Code

claude plugin marketplace add psyb0t/agents
claude plugin install proxq@psyb0t

Claude Code asks for the proxq base URL (PROXQ_URL) when the plugin is enabled.

Codex

codex plugin marketplace add psyb0t/agents
codex plugin add proxq@psyb0t

Installed through the marketplace, the skill is invoked as $proxq:proxq. Codex also picks the skill up with no install in any repository that contains .agents/skills/, where it is invoked as plain $proxq.

OpenClaw

The skill is published to ClawHub on every release:

openclaw skills install @psyb0t/proxq

What the agent is told

The skill carries the same warnings as the security page: proxq has no authentication, it makes outbound requests on the caller’s behalf, and there is no ownership check on job IDs, so the agent cancels only jobs it submitted or that you named.