Installs into .claude/skills of the current project.
Are you the author of Speckit Jit Branch?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/iamalisson-speckit-jit-branch)
---
name: speckit-jit-branch
description: Switch to the task's branch before implementing
compatibility: Requires spec-kit project structure with .specify/ directory
metadata:
author: iamalisson
source: jit:commands/branch.md
---
# Task branch (before `/speckit-implement`)
One task = one issue = one branch = one PR. This puts the session on the task's branch before
any code is touched.
1. Find the task ID: the single `T###` in the arguments passed to `/speckit-implement`.
- If there is none, or more than one: **stop the whole implement run**. Tell the user to pass
exactly one task, e.g. `/speckit-implement T003`. Do not implement anything.
- A feature number or folder in the arguments (`002`, `002-reports`) names the feature; without
it, the active one is used. `/speckit-implement 002 T001` is T001 of feature 002.
2. Run `bash .specify/extensions/jit/scripts/branch.sh [<FEATURE>] <TASK_ID>`. It also pushes a
new branch to origin at once, so the task shows as in progress to every clone and worktree.
- On failure, stop the implement run and show the script's message as is. It explains what is
missing: the issue, the spec PR, the analysis, or a clean working tree. Never work around
it: no stash, no reset, no checkout of other files.
3. On success, print `On <BRANCH> for <TASK> (#<ISSUE>)` and continue `/speckit-implement`
with **only that task**. Follow the constitution's defaults while implementing.
4. Do not tick the task's checkbox in `tasks.md`. The task is done when its issue closes.
