Recurring Task Skill
Guidelines for executing recurring autonomous tasks effectively.
Execution Pattern
- Review history: Check your conversation history for outcomes from previous runs
- Check state: Use
memory_readto recall any notes from last time - Execute goal: Use tools to accomplish the task
- Verify: Confirm the action succeeded (check API responses, file changes, etc.)
- Log outcomes: Use
memory_logto record what happened
Error Recovery
- If credentials are missing, log which credentials are needed and where to put them
- If an API is down, log the failure and skip gracefully — don't retry endlessly
- If a task partially succeeds, log what worked and what didn't
State Between Runs
Use memory_log to persist information across runs:
memory_log: "daily-report: Posted successfully. Engagement on yesterday's post: 12 likes, 3 replies."
memory_log: "repo-check: 2 new issues found, CI green on all repos."
Credential Management
Check these locations for API keys:
- Environment variables (e.g.,
$GITHUB_TOKEN,$SLACK_WEBHOOK) - Config files (
~/.cereworker/config.yml) - Credential stores (
~/.config/gh/hosts.ymlfor GitHub)
If credentials are missing, clearly describe what's needed in the task output.