Task Centered Workflow Management With Structured Prds And State Tracking

via “task-centered workflow management with structured prds and state tracking”

The best agent harness.

Unique: Implements task lifecycle as a first-class concept with task.json state files and task.py scripts, enabling AI agents to understand and update task progress programmatically. Tasks are version-controlled and archived, creating an audit trail of AI-assisted work with explicit scope and dependencies.

vs others: Unlike GitHub Issues or Jira, Trellis tasks are embedded in the codebase (.trellis/tasks/) and designed for AI agent consumption, with structured PRDs and state files that agents can read and update directly. Unlike linear task runners, Trellis integrates task context into AI sessions automatically via context injection.