From claude-code-hermit
Stops Docker container running autonomous hermit agent, marks session as operator takeover, loads full context from config and session files, and presents structured summary for interactive continuation.
npx claudepluginhub gtapps/claude-code-hermit --plugin claude-code-homeassistant-hermitThis skill uses the workspace's default tool permissions.
Stop the autonomous hermit and take the wheel. This skill is for Docker-based always-on setups — it stops the container, preserves all session state, and loads the hermit's context so you can work interactively with full continuity.
Summarizes operator takeover activity via git log, optionally queues instructions in NEXT-TASK.md, updates SHELL.md status, and restarts hermit Docker container after /hermit-takeover.
Troubleshoots Hermes Agent ops issues like gateway failures, platform drops, cron non-execution, profile config chaos, and inactive skills/tools via hermes CLI commands in check-locate-fix-verify steps.
Manages multi-session autonomous agent tasks with progress checkpointing, failure recovery, task dependencies, and commands like /harness init/run/status/add. For long-running workflows across context windows.
Share bugs, ideas, or general feedback.
Stop the autonomous hermit and take the wheel. This skill is for Docker-based always-on setups — it stops the container, preserves all session state, and loads the hermit's context so you can work interactively with full continuity.
Read .claude-code-hermit/config.json to get:
tmux_session_name — resolve {project_name} with the actual project directory nameagent_name — for displayRun docker compose -f docker-compose.hermit.yml ps --status running --format '{{.Name}}' in the project root.
docker compose fails (Docker not installed or no compose file): inform the operator: "No docker-compose.hermit.yml found. This skill is for Docker-based setups." and stop.Run docker compose -f docker-compose.hermit.yml stop (not down — we want to preserve the container for restart).
This sends SIGTERM, which triggers the entrypoint to exit cleanly. If hermit-stop runs inside the container, it archives the current report before exiting.
Wait briefly for the container to stop: check docker compose -f docker-compose.hermit.yml ps --status running up to 3 times with 5-second intervals. If it's still running after 15 seconds, warn: "Container is still stopping — it may take a moment for the session to archive."
Read .claude-code-hermit/sessions/SHELL.md.
If the file exists and has content:
**Status:** line and change the value to operator_takeover- **Takeover:** YYYY-MM-DD HH:MM with the current timestampIf SHELL.md doesn't exist or is empty: note this for the summary — the hermit may have been idle with no active session.
Read these files (all in parallel):
.claude-code-hermit/OPERATOR.md — the project rulebook.claude-code-hermit/sessions/SHELL.md — current/last session stateS-*-REPORT.md in .claude-code-hermit/sessions/ (by filename sort, descending)Print a structured summary for the operator:
Takeover complete.
Session: S-NNN | was: in_progress
Task: "Add input validation to API endpoints"
Progress: 2/4 tasks completed (via TaskList)
Blockers: none
Cost so far: $1.80
You're now driving. The hermit's full context is loaded.
When you're done, run /claude-code-hermit:hermit-hand-back to restart the hermit.
Adapt based on what's actually in SHELL.md:
The session continues normally from here. The operator works interactively with the full hermit context loaded.