Hand off tasks between AI sessions
Leave a structured handoff in a local task database so the next session can tell what was finished, what remains and what needs review.
Hand off tasks between AI sessions when checking the work will take longer than the current conversation. The td-task-management skill records issue state and progress in a project-local task store. A structured handoff can distinguish completed work, remaining actions, decisions and uncertainties.
A team changing a reporting script may have finished the new output but still need to check older files. Record the compatibility check as remaining. The next session can inspect the issue and the previous reasoning before modifying anything. Submitting the issue for review is a separate step from approving that review.
What it does
- Record an outcome rather than a vague percentage complete.
- Keep the reasoning for a decision beside the affected issue.
- Group related tasks while keeping each issue’s state visible.
- Distinguish an unresolved question from a technical blocker.
- Submit completed work for review and leave an independent reviewer a clear trail.
Install and use
- Use the original td CLI and give the agent access to the intended project’s task store.
- Find the issue and read its status and recent handoff before starting new work.
- Record progress and outstanding checks, then submit for review when appropriate.
FAQ
Source
Made by Marcus Vorwaller. Original skill: td-task-management, published on GitHub under the MIT licence.
View on GitHub