finish-branch
Wrap up the current branch or worktree: clean up debugging leftovers, run final verification, then let the user decide whether to merge, keep, open a PR, or…
Skill by RollRollRoll
Chalk Labs
Need help deploying finish-branch?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/rollrollroll-finish-branch
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/finish-branch
What's inside
- Skill name
finish-branch- Description
- Wrap up the current branch or worktree: clean up debugging leftovers, run final verification, then let the user decide whether to merge, keep, open a PR, or discard.
- License
- Apache-2.0
- Instructions
- 92 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- openai.yaml · agents/
SKILL.md (first lines)
# Finish Branch: Wrapping Up Development ## Goal After a stretch of development is **implemented**, **wrap up** the current branch / worktree: first run a read-only state preflight, then clean up debugging leftovers and run the final tests, and finally offer the options "merge / keep locally / create PR / discard", leaving destructive and external actions to the user. Use it after development is done, to clean up leftovers, run final verification, and decide where the branch goes. Setting up a workspace is not its job. ## Invocation contract **Inputs** - **Final verification command**: how to run the project's full acceptance check. - **Scope of this development** (optional): used to judge which debugging leftovers were introduced by this work; if not given, clean only what can be confirmed. - **Caller** (optional): when called by another skill, state the caller. If the cleanup produces code changes, control must go back to the caller to redo the overall acceptance check. **Returns** - Preflight conclusion (branch / worktree / staged and untracked state, ahead/behind relative to the main branch). - What was cleaned up, final test results, the wrap-up option the user chose, and **the actions actually taken**. - For an unauthorized action, stop before it and report truthfully; do not decide for the user. **Responsibility boundaries** - Do not wrap up if tests fail. Do not automatically commit / merge / push / open a PR / delete a branch / remove a worktree; get confirmation for each one separately. - Do not automatically stash, `git clean`, or silently discard to create a "clean state". - Do not set up a workspace or do code review; this skill only wraps up. ## When to use / when not to use **Use it when**: - A stretch of development is done and tests pass, and you need to decide what to do with the branch. - Wrapping up after the overall acceptance check passes. - You finished work in a worktree and want to merge it back and clean up the worktree. **Do not use it when**: - Development is not finished or tests fail: finish development and fix the tests first, then wrap up. - You need to set up an isolated workspace: that is not this skill's job. ## Core flow (state preflight → clean up debugging leftovers → final verification → offer options → execute with authorization) 1. **Run a read-only state preflight first**: - Confirm the current branch, whether you are in a worktree, and staged / unstaged / untracked state. - If there are any non-ignored changes on entry, show a summary and stop, and let the user decide. **Do not clean up, commit, stash, or discard first**, so user changes do not get mixed into the wrap-up. 2. **Clean up debugging leftovers from this development**: - Check this diff for temporary logs, debug prints, breakpoints, temporary flags, and code added only to investigate a problem. - Clean up only leftovers you can confirm this development introduced. When the purpose is unclear, ask the user first; do not touch unrelated code along the way. - Cleanup must happen before the final tests, so a "green before cleanup" result is not taken as the final result. 3. **Run the final tests and confirm all green**: run the full acceptance check the project defines. **Not green, no wrap-up**: go back and fix, then rerun the final tests. 4. **Recheck repo / worktree state**: - If cleanup produced changes, show the diff and get explicit commit authorization before committing; after committing, confirm the state is clean again. - If cleanup produced code changes, control must go back to the caller to redo the overall acceptance check; the review conclusion from before cleanup cannot be reused. - Relationship to the main branch (how many commits ahead / behind). 5. **Offer the wrap-up options**: - **Merge to main**: merge back into the main branch. - **Keep locally**: leave it on the branch, do not merge yet. - **Create PR**: push and open a PR; write the PR body by invoking `pr` with the Skill tool. - **Discard**: abandon this work (delete the branch / worktree). **Irreversible**. 6. **Execute after getting the matching authorization**: - `git commit`, merge, push, creating a PR, and deleting a branch all need the user's explicit confirmation of the specific action before running. - Even if the code is merged or discarding is decided, `git worktree remove` needs separate confirmation; do not read "chose merge" as also authorizing removal of the worktree. - For keep locally, only report the branch / worktree path and state; do not make implicit commits or cleanups. ## Safety gates - **⚠️ State changes need explicit user confirmation, never automatic**: commit, merge, push, create PR, delete branch, discard, remove worktree. Follow the current repo's `CLAUDE.md` / `AGENTS.md` and the platform's approval policy. - **No wrap-up if tests fail**: go back and get them green first. - **Tests must come after debug cleanup**: any cleanup or fix invalidates earlier test results, so rerun them. - **The overall acceptance check also goes stale**: if cleanup produces code changes after the overall acceptance check passed, review again. - **Discarding is irreversible**: **confirm twice** before deleting a branch; also get explicit confirmation before removing a worktree. - **Detect before acting**: acting before the state is clear (wrong branch, missed uncommitted changes) can cause damage. - **No implicit fallbacks**: do not use automatic stash, automatic commits, `git clean`, or silent discards to create a "clean state". > Check: does every commit, merge, push, PR, deletion, or worktree removal you are about to run have explicit authorization for that action? If one is missing, stop before it. ## Common mistakes - Wrapping up / merging when tests fail. - Running tests first, deleting debug prints afterwards, and not rerunning tests. - Starting cleanup when there were dirty changes on entry, mixing user changes into the wrap-up. - Changing code after the overall acceptance check and then merging on the old review conclusion. - Finding uncommitted changes and automatically committing / stashing to keep the flow going. - Automatically merging / pushing / opening a PR without letting the user decide (dangerous operation). - Removing the worktree after the merge is approved: the user only confirmed the merge, not the removal. - Discarding (deleting the branch) without a second confirmation. - Acting without checking state: on the wrong branch / missing uncommitted changes. - Setting up a workspace here: out of scope, this skill only wraps up.
From RollRollRoll/agent-toolkit, Apache-2.0.
More from RollRollRoll/agent-toolkit
- 0.570score
tdd
RollRollRoll
Build features or fix bugs test-first.
SkillAFreshness: AClaude Code +9 - 0.570score
diagnosing-bugs
RollRollRoll
Diagnose hard bugs and performance regressions with a reproducible, verifiable loop.
SkillAFreshness: AClaude Code +9 - 0.570score
codebase-design
RollRollRoll
Design deep modules with a shared vocabulary.
SkillAFreshness: AClaude Code +9 - 0.567score
git-commit-push
RollRollRoll
Resolve an in-progress git merge, rebase, or cherry-pick conflict, or organize finished changes into atomic commits and push them after the user confirms.
SkillAFreshness: AClaude Code +9 - 0.567score
codebase-analyzer
RollRollRoll
Investigate an unfamiliar project or codebase and explain its purpose, architecture, and implementation in a structured Chinese report.
SkillAFreshness: AClaude Code +9 - 0.561score
grill-demand
RollRollRoll
Interview the user in depth to settle problem, scope, behavior, and technical decisions layer by layer, creating docs (ADRs and glossary) as you go.
SkillAFreshness: AClaude Code +9 - 0.558score
setup-worktree
RollRollRoll
Create and verify an isolated Git worktree, locking the base commit explicitly before creating it.
SkillAFreshness: AClaude Code +9 - 0.555score
implement-spec
RollRollRoll
Implement the result of to-spec and to-tickets in code.
SkillAFreshness: AClaude Code +9 - 0.544score
domain-modeling
RollRollRoll
Build and sharpen a project's domain model.
SkillAFreshness: AClaude Code +9 - 0.544score
prototype
RollRollRoll
Build a throwaway prototype to answer a design question.
SkillAFreshness: AClaude Code +9 - 0.541score
triage
RollRollRoll
Move issues and external PRs through a state machine of triage roles, categorise, verify, grill if needed, and write agent-ready briefs.
SkillAFreshness: AClaude Code +9 - 0.541score
setup-env
RollRollRoll
Configure this repo for these skills: set up its issue tracker, triage label vocabulary, and domain doc layout.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 19, 2026 (12 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/RollRollRoll/agent-toolkit
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from RollRollRoll
- 0.570score
codebase-design
RollRollRoll
Design deep modules with a shared vocabulary.
SkillAFreshness: AClaude Code +9 - 0.570score
tdd
RollRollRoll
Build features or fix bugs test-first.
SkillAFreshness: AClaude Code +9 - 0.570score
diagnosing-bugs
RollRollRoll
Diagnose hard bugs and performance regressions with a reproducible, verifiable loop.
SkillAFreshness: AClaude Code +9 - 0.567score
git-commit-push
RollRollRoll
Resolve an in-progress git merge, rebase, or cherry-pick conflict, or organize finished changes into atomic commits and push them after the user confirms.
SkillAFreshness: AClaude Code +9 - 0.567score
codebase-analyzer
RollRollRoll
Investigate an unfamiliar project or codebase and explain its purpose, architecture, and implementation in a structured Chinese report.
SkillAFreshness: AClaude Code +9 - 0.561score
grill-demand
RollRollRoll
Interview the user in depth to settle problem, scope, behavior, and technical decisions layer by layer, creating docs (ADRs and glossary) as you go.
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get finish-branch running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
finish-branch skill. Agent Mart. https://agentmart.dev/skills/rollrollroll-finish-branch (last updated 2026-10-24)