OpenClaw · Skill
Sport Mode
Temporarily boost heartbeat frequency (default 3m) and inject a monitoring task into HEARTBEAT.md . Perfect for supervising background agents (Codex), long-running builds, or interactive games.
Install
Start with the primary install command. Alternate entrypoints are included below for ClawHub and OpenClaw CLI users.
Primary command
clawhub install l1vein/sport-modeClawHub installer
npx clawhub@latest install l1vein/sport-modeOpenClaw CLI
openclaw skills install l1vein/sport-modeDirect OpenClaw install
openclaw install l1vein/sport-modeWhat this skill does
Temporarily boost heartbeat frequency (default 3m) and inject a monitoring task into HEARTBEAT.md . Perfect for supervising background agents (Codex), long-running builds, or interactive games.
Why it matters
Patching the config at runtime triggers a live Gateway reload, so you get faster heartbeats without restarting the agent process.
Typical use cases
- Watching a Codex agent complete a multi-file refactor
- Monitoring a long database migration until it finishes
- Tracking CI build progress at 1-minute intervals
- Running a turn-based game loop with per-tick state updates
- Supervising a staged deployment with per-stage status checks
Source instructions
Sport Mode
Temporarily boost heartbeat frequency (default 3m) and inject a monitoring task into HEARTBEAT.md.
Perfect for supervising background agents (Codex), long-running builds, or interactive games.
Usage
# Turn ON: Set heartbeat to 3m and set monitoring task
sport-mode on --task "Check Codex progress. If done, run sport-mode off."
# Custom Interval: Set to 1 minute
sport-mode on --task "Game tick" --every "1m"
# Turn OFF: Reset heartbeat to 30m and clear HEARTBEAT.md
sport-mode off
How it works
- ON:
- Patches
~/.openclaw/openclaw.json(hot-reload) to setheartbeat.every. - Writes your task to
HEARTBEAT.mdwith a "Sport Mode Active" header.
- Patches
- OFF:
- Patches config back to
30m(default). - Clears
HEARTBEAT.md.
- Patches config back to
Best Practices
1. Set a Finish Line
Unless you want an endless marathon, always define a termination condition in your task.
- ✅ Good: "Monitor build. If success or fail, run sport-mode off."
- ❌ Bad: "Monitor build." (Agent might keep reporting "Done" forever until you manually stop it).
2. State Machine in File
For multi-step tasks (like games or staged deployments), let the agent update HEARTBEAT.md itself.
- Pattern: Read state -> Execute step -> Write new state -> Sleep.
- This keeps the agent "stateless" (doesn't rely on conversation history context window) but the task "stateful".
3. Use tmux for Visibility
If the monitoring task involves terminal output (e.g., Codex coding, compiling), running the task in a tmux session is ideal.
- The agent can inspect the pane (
tmux capture-pane) without interfering. - The user can attach (
tmux attach) to watch live.
4. Silence is Golden
For high-frequency modes (e.g., 1m), avoid spamming "Nothing happened".
- Configure the agent to reply
HEARTBEAT_OK(silence) if the status hasn't changed. - Only notify the user on milestone completion, errors, or final success.
Implementation Note
This skill uses openclaw config set to safely patch configuration at runtime, triggering a seamless Gateway reload.