Skip to content

Taskwarrior Workflow — Project Manager

Use Taskwarrior to publish ready queues per agent, enforce dependency order, and shard the backlog when multiple agents work in parallel.

Steps

  1. Ingest sources: Refresh tasks from docs/features/**, docs/issues/**, docs/workplan/**; update priorities/dependencies.
  2. Export/Backfill:
  3. Export current Taskwarrior state: just taskwarrior-export
  4. Backfill from JSON (dry-run): just taskwarrior-backfill
  5. Apply backfill: just taskwarrior-backfill-apply
  6. Shard pools (when N agents run in parallel):
  7. Partition tasks into N pools by role/priority: just taskwarrior-pools pools=<N> (writes outputs/workplan/taskwarrior-pool-*.json).
  8. Ensure no cross-pool dependency; reassign or reorder until pools are independent.
  9. Publish ready queues:
  10. For each pool/role: provide the ready filter (task +role:<ROLE> status:pending -BLOCKED next) or just taskwarrior-next role=<ROLE>.
  11. Export per-pool JSON for clarity (see pool outputs).
  12. Monitor & adjust:
  13. Rebalance pools if one agent drains early.
  14. Update priorities/dates based on operator feedback and dependency changes.

Notes

  • Keep pools dependency-free; if impossible, merge affected tasks into the same pool.
  • Annotate tasks with source doc path and target date to maintain traceability.
  • Provide agents a short “next up” list per role after each planning cycle.