to-do

What This Skill Does

The To-Do skill transforms your OpenClaw agent from a reactive assistant into a proactive partner. By leveraging native OS scheduling mechanisms (at/schtasks), it allows you to offload tasks to the future. Unlike simple reminders, this skill performs 'context injection,' meaning when the timer fires, the agent wakes up with the specific instructions and environmental parameters required to execute complex workflows, system checks, or notifications without your manual intervention.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/devlumuz/to-do

Ensure you have configured the required environment variables: OPENCLAW_BIN (the absolute path to your binary) and OPENCLAW_TZ (your IANA timezone, e.g., 'America/New_York'). Without these variables, the skill cannot map your local time to the system scheduler, and it will fail to initialize.

Use Cases

• Automated Reporting: Schedule daily, one-off system health reports to be generated and sent to your communication channels at a specific hour.
• Delayed Task Hand-offs: If you need to remind the agent to follow up on a project later in the week, it can wake itself up to perform the check.
• Workflow Automation: Schedule complex data migrations or backups to run during off-peak hours by setting a precise execution time.
• Deadline Management: Use it to nudge the agent to check the status of pending items in your project management tools automatically.

Example Prompts

1. "Check the status of the server migration project in my documents folder tomorrow at 9:00 AM and notify me if the build logs show an error."
2. "At 5:00 PM today, scan the active directory and send a summary of all modified files to the team Slack channel."
3. "Remind me to finalize the Q3 marketing report by listing all pending tasks in the current directory on Friday at 2:00 PM."

Tips & Limitations

• Always run 'now' first: Never assume the agent knows what time it is. The 'now' command is your single source of truth for relative time.
• Avoid Ambiguity: Because the agent wakes up with an isolated context, your instructions must be explicit. Avoid pronouns like 'he' or 'it'—refer to specific file paths, full names, and exact URLs.
• Security: The scheduler strictly prohibits shell meta-characters. This is a deliberate design choice to prevent command injection. Keep your instructions to plain, clear text.