This page provides a low-tech task board that works anywhere because it’s just a file.
Download
- Template file:
/templates/openclaw/TASKS.md
Direct link:
Install into your OpenClaw workspace
- Confirm the workspace path used by your running gateway:
openclaw config get agents.defaults.workspace
- Copy the template into that directory (example):
cp TASKS.md ~/.openclaw/workspace/TASKS.md
If you’re on Docker, remember the workspace path inside the container may differ; validate with config/volume mounts:
The one rule: every run leaves evidence
For cron/heartbeat runs:
- Write a timestamped artifact into workspace (report markdown/JSON).
- Append one line into Run Log with: timestamp, task id, OK/FAIL, artifact path.
Cron reliability guide:
Plug-and-play: ready-to-paste cron prompts
Use these as your cron job prompt content. They are designed to work even if delivery is flaky because they always:
- write a report artifact under
reports/ - append a single line to
TASKS.mdRun Log
Tip: for maximum reliability, set cron delivery to none and explicitly send the summary message yourself (see the cron troubleshooting pages linked above). If you keep delivery as announce, the artifact + Run Log still gives you proof.
Prompt A: Daily “system health + model probe” report (safe baseline)
Replace T-HEALTH with your task id.
You are running as a scheduled cron job.
Goal: produce a daily health report for this OpenClaw instance and leave durable evidence.
Steps:
1) Determine current timestamp in ISO format (YYYY-MM-DD and time).
2) Run a minimal internal sanity check:
- Confirm gateway status (deep).
- Confirm the default model is callable (probe).
- Confirm channels are reachable (probe).
3) Write a report file to: reports/health/YYYY-MM-DD.md
The report must include:
- timestamp
- a short OK/FAIL summary for each probe
- any error snippets (redacted; do not include tokens)
4) Append one line to TASKS.md under Run Log:
`YYYY-MM-DD HH:MM` | `T-HEALTH` | `OK|FAIL` | `cron:health` | `reports/health/YYYY-MM-DD.md` | `<short error summary or ->`
5) If delivery is enabled for this cron, send a short message:
- “Health report OK/FAIL — see reports/health/YYYY-MM-DD.md”
Constraints:
- Do NOT paste secrets into the report or TASKS.md.
- If any step fails, still write the report file with failure details and mark the Run Log line as FAIL.
Prompt B: Daily browser automation “smoke test” (leaves evidence)
Replace T-BROWSER and the URL.
You are running as a scheduled cron job.
Goal: run a tiny browser automation smoke test and leave evidence for debugging timeouts.
Target:
- URL: https://example.com
Steps:
1) Determine current timestamp in ISO format (YYYY-MM-DD and time).
2) Use the browser tool to open the target URL and extract:
- page title
- first 200 characters of visible body text (no secrets)
3) Write a report file to: reports/browser-smoke/YYYY-MM-DD.md
Include:
- timestamp
- URL
- result (OK/FAIL)
- extracted title/text (if OK)
- error summary (if FAIL)
4) Append one line to TASKS.md under Run Log:
`YYYY-MM-DD HH:MM` | `T-BROWSER` | `OK|FAIL` | `cron:browser-smoke` | `reports/browser-smoke/YYYY-MM-DD.md` | `<short error summary or ->`
5) If the run fails with a browser timeout, include the exact error string in the report.
6) If delivery is enabled for this cron, send a short message:
- “Browser smoke OK/FAIL — see reports/browser-smoke/YYYY-MM-DD.md”
Constraints:
- Fail fast: do not loop endlessly. One attempt is enough for a smoke test.
- Do NOT paste cookies, tokens, or any sensitive page data into the report.
Related:
- Browser timeout troubleshooting: /guides/openclaw-browser-automation-timeouts
- “browser control service timeout” fix: /troubleshooting/solutions/browser-control-service-timeout
Safety notes
- Don’t paste tokens/API keys into TASKS.md.
- Don’t paste full sensitive emails; link to redacted artifacts instead.
Security playbook: