run-agent — Claude Code skill: launch agents via CCAM
This skill starts Claude or Claude runs through the local CCAM Run API and keeps an eye on them from the chat.
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a claude-codeWhat run-agent does
run-agent sits on top of the local CCAM dashboard and hands the agent a set of ccam run commands. Before anything starts, it checks the provider binary with ccam run binary claude or ccam run binary codex, lists what models are available via ccam run models <provider>, and confirms the target folder with ccam run cwds. Then it prints the exact launch settings — provider, prompt, working directory, model, approval mode, sandbox, resume session ID — and waits for your go-ahead.
Once a run is live, ccam run list and ccam run get <id> --envelopes show the output as it comes, ccam run send <id> --text "..." pushes a follow-up, and ccam run stop <id> --yes ends it. Keep the provider the same as the one that started the run when you send a message; switching it breaks the session.
The skill steers away from danger-full-access unless you ask for it outright. Nothing gets started, messaged or stopped without confirmation. Run history counts as evidence here: if the final output or the saved session reports an error, process status alone is not proof the work finished.
Install it with npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a claude-code; the skill file lands at ~/.claude/skills/run-agent/SKILL.md.
What it helps with
- start a Claude or Claude run that the dashboard can watch
- pick the model, approval policy and sandbox before launch
- read live output without leaving the terminal
- send a follow-up into a session that is already running
- stop a launched run with one command
Example prompts for the agent
- Check the Claude binary and list the models it supports
- Start a run in
/path/to/repowith the prompt "Review the current changes" - Show me the live output and send a follow-up to that run
How to install run-agent
1Skills CLI (Vercel)
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a claude-code2Manually: put the files in place
~/.claude/skills/run-agent/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/run-agent/
1Skills CLI (Vercel)
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a codex2Manually: put the files in place
.agents/skills/run-agent/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a cursor2Manually: put the files in place
.cursor/skills/run-agent/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install hoangsonww/Claude-Code-Agent-Monitor run-agent2Manually: put the files in place
.github/skills/run-agent/SKILL.md1Skills CLI (Vercel)
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a gemini-cli2Platform CLI
gemini skills install https://github.com/hoangsonww/Claude-Code-Agent-Monitor --path plugins/ccam-runner/skills/run-agentManually: put the files in place
.devin/skills/run-agent/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about run-agent
How do I install run-agent in Claude Code?
Run npx skills add hoangsonww/Claude-Code-Agent-Monitor --skill run-agent -a claude-code. The skill file goes to ~/.claude/skills/run-agent/SKILL.md.
What does run-agent do?
It launches and supervises Claude or Claude through the local CCAM Run API: choosing the model, sandbox and working directory, sending follow-ups, reading live output and stopping the run.
Which access mode does the skill use?
It avoids danger-full-access and only applies it when the user asks for it explicitly. Starting, messaging or stopping a run all require confirmation.
Source text
SKILL.md · MIT
--- name: run-agent description: > Launch and supervise Claude Code or Codex through the CCAM Run API. Use when the user wants to start a monitored agent, select a model, approval policy, sandbox, or working directory, send a follow-up, inspect live output, resume a native session, or stop a dashboard-launched run. --- # Run Agent Use `ccam run` against the local dashboard. ## Workflow 1. Verify the provider binary: - `ccam run binary claude` - `ccam run binary codex` 2. Discover supported models with `ccam run models <provider>`. 3. Confirm the working directory with `ccam run cwds`. 4. Show the exact launch settings before starting. Include provider, prompt, working directory, model, approval mode, sandbox, and resume session ID. 5. Start only after user confirmation: ```bash ccam run start --provider codex --cwd /path/to/repo \ --prompt "Review the current changes" \ --permission on-request --sandbox workspace-write --yes ``` 6. Inspect with `ccam run list` or `ccam run get <id> --envelopes`. 7. Send follow-ups with `ccam run send <id> --text "..." --provider codex --yes`. 8. Stop with `ccam run stop <id> --yes`. ## Safety - Never use `danger-full-access` unless the user explicitly requests it. - Do not start, message, or stop a run without confirmation. - Preserve the provider used to start the run when sending messages. - Treat run history as evidence. Do not claim completion from process status alone when the final output or persisted session shows an error.
© hoangsonww · MIT · hoangsonww/Claude-Code-Agent-Monitor
