claude-for-agentslisted

# Claude for Agents ## Overview Claude for Agents turns the locally authenticated official Claude Code CLI into a persistent, permission-controlled MCP worker. A compatible supervisor such as Hermes or Codex can delegate work, receive structured telemetry, and continue the exact Claude session later. Use this skill to install, update, configure, verify, operate, or troubleshoot the bridge. The runtime is installed from its GitHub source; it is not an npm package. ## When to Use Use when the user wants to: - add Claude Code as a worker beneath Hermes, Codex, or another stdio MCP host; - delegate implementation or read-only review work to Claude; - resume an exact Claude worker session; - inspect or cancel background Claude jobs; - update or troubleshoot an existing Claude for Agents installation. Do not use this skill to: - bypass Claude authentication, subscription limits, or provider controls; - expose the bridge as a public or remote service; - send secrets through prompts; - treat an MCP configuration entry as proof that live tool discovery works. ## Safety Contract 1. **Local auth only.** Never ask for, read, print, copy, or type Claude credentials. Check `claude auth status`; if authentication is missing, tell the user to run `claude auth login` themselves and stop before live delegation. 2. **Approval before changes.** Explain before cloning, downloading dependencies, changing MCP configuration, or restarting an agent. Do not silently replace an existing MCP e