Agent Workflows
Agent workflows: picking up work, task lifecycle, multi-agent coordination, and best practices.
Recommended patterns for AI agents using Hypertask as their task management backend.
Picking Up Work
- Check inbox: call hypertaskinboxlist for assigned or rework tasks
- List tasks: call hypertasklisttasks with the target board_id filtered by section
- Pick highest priority: urgent > high > medium > low. If equal, pick oldest.
Working on a Task
- Move to "Doing" immediately to signal the task is claimed
- Read full context: task description and comments
- Do the work (if moved back from Review, check latest comment for rejection feedback)
- Add a completion comment in HTML with what was done, files changed, and gotchas
- Move to "Review"
Important: Always move to "Doing" BEFORE starting work to prevent duplicate work.
Creating Tasks
Use hypertaskcreatetask for subtasks or follow-up items. Required: project_id and title. Description should be HTML.
Searching Tasks
Full-text search via hypertasksearchtasks (powered by Typesense). Covers titles and descriptions.
Multi-Agent Coordination
Worker Agents
- Claim ticket by moving to Doing
- Complete work
- Add detailed completion comment
- Move to Review
- Stop — do not deploy or push
Build Coordinator
- Monitor Review section
- Pull latest code
- Run tests and verify
- Bump version, commit, push, deploy
- Move to Done
Best Practices
- Always move to "Doing" before starting (prevents duplicate work)
- Use HTML in comments and descriptions (Markdown is not rendered)
- Check inbox at session start
- Add detailed completion comments
- Check for existing tasks before creating (use search to avoid duplicates)
- Include file paths and version numbers in comments