Plan, prove and hand off a task with the CLI
You see the whole loop once: the plan, the claim, the test evidence and the handoff all travel with the repository.
- SpecWeave CLI
- 1:00
- Any agent
- Narrated, with captions
Recorded on SpecWeave 3.0.6, added 12 October 2026.
Before you start: Node.js 20.12 or later and git; SpecWeave installed with npm i -g specweave.
Steps
- Create an increment with
specweave create-increment "title", then write the problem, the acceptance criteria and a task with its exact files and test command in spec.md. - Claim the task with
specweave task claim T-01, so everyone can see who owns it. - Implement the change and run the tests. In the recording, both tests pass.
- Commit, then run
specweave task done T-01 --run "<test command>". SpecWeave runs the command again and keeps its output as evidence. - Stop with
specweave handoff --reason "<why>". The recording keeps everything local instead of pushing. - In another clone, run
specweave pickup. It prints the next task with its acceptance criteria.
Transcript
Built from the caption file of the narration.
0:00 A disposable demo repository
Here is a real SpecWeave workflow in a disposable demo repository. No production code is touched.
0:07 Create an increment
First, create an increment. Write down the problem, the acceptance criteria, and a task with its exact files and test command.
0:15 Claim the task
Now claim the task, so ownership is visible.
0:18 Implement and test
Implement the function and run the tests. In this example, we check a list of numbers and an empty list. Both tests pass.
0:25 Commit and mark done
Commit the implementation. Then mark the task done with the test command. SpecWeave runs it again and records the evidence.
0:33 Hand off
To stop, create an explicit handoff. This demo uses no push, so everything stays local. We save that handoff and make a clean local clone.
0:43 Pick up in a clean clone
Run pickup in the clone. The next task is still there: document the empty-list behavior. The plan, ownership, and evidence travel with the repository. Try the same workflow in your own small project at spec weave dot com.
Learning path
Run work across agents
Plan an increment, prove each task, hand off between tools and get a review from another model.
Plan, prove and hand off a task with the CLI (this guide)
1:00, SpecWeave CLI
Install SpecWeave and plan your first increment
1:36, SpecWeave CLI
Claim, prove, verify and complete a task
1:30, SpecWeave CLI
Hand off from Claude Code to Codex at a usage limit
1:18, SpecWeave CLI
Review with the other model before you ship
1:33, SpecWeave CLI