Ticket dependencies
You can mark a ticket as blocked by one or more other tickets. Trundle uses this to sequence agent work:
- A dependent ticket will not be picked up by the Trundle Dev Agent until every ticket it depends on has actually been merged: that is, the blocker is completed and its pull request has been merged into the base branch.
- Once its dependencies are merged, the dependent ticket branches directly off the base branch, so it builds on the already-merged code cleanly. There is no synthetic branch stacking, and nothing needs to be "extracted" afterwards.
If you drag a dependent ticket into In Progress before its blockers are merged, Trundle holds it until they're ready.
Fail-safe assignment tracking
Tickets never get lost when an agent run hits a problem:
- Unassigned ticket → agent column: If you drag an unassigned ticket into an agent-automated column, Trundle records you as the ticket's previous assignee and assigns the run to the system user.
- On failure: If the agent run fails, the ticket moves back to the Backlog, is returned to that person (falling back to the reporter), and an Action Item is created so the handoff isn't missed. For a ticket that was already assigned, the original assignee is restored instead.
Watching the agents work
While the Trundle Dev Agent or Review Agent is working on a ticket, the Thought process tab shows live progress:
- a "Trundling…" activity pill while a dev or review run is active, and
- a live typing animation as the agent streams its reasoning.
(See the Boards & Backlog and QA articles for how tickets flow through the rest of the pipeline.)