sidequest
Skill1 starsChanged 7 months agoMITView on GitHub
Run a task in a background git worktree that opens a PR, so your main checkout stays free to keep working.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- kevmoo/kevmoo_skillstheir own version, first seen Nov 5, 2026
- shahrukh-hack/vibe-skillstheir own version, first seen Oct 20, 2026
- michaelboeding/skillstheir own version, first seen Jan 5, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- michaelboeding/skills★ 30same name
- kevmoo/kevmoo_skills★ 22same name
- shahrukh-hack/vibe-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 24
If the worker reports STATUS: isolation-failed on its first verification step and the status file is empty, a missing permission pattern is …
- Distinctive: line 45
Call the Agent tool with:
- Very specific: line 63
Do NOT pass isolation: "worktree".
- Distinctive: line 93
Do not do anything else. The loop skill owns the rest.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
You are a dispatcher. Your job is to explicitly create an isolated worktree, launch a background agent inside it, then hand off to /loop so you can tail the worker's status file and report progress as it writes.
⚠️ Prerequisites — required settings.json permissions
The worker runs as a background sub-agent. Background sub-agents CANNOT prompt the user for bash permission, so any command not pre-approved in ~/.claude/settings.json (or project .claude/settings.local.json) silently fails. The following patterns MUST exist in permissions.allow before /sidequest can succeed:
Bash(git worktree add:*)Bash(git worktree remove:*)Bash(git -C:*)Bash(yarn --cwd:*)(orBash(npm --prefix:*)/Bash(pnpm --dir:*)/Bash(pnpm -C:*)for that package manager)Bash(gh -R:*)Bash(gh repo view:*)
Notes:
- Settings changes may require a Claude Code session restart to take effect. If the worker was launched in a session that started before these patterns were added, it won't see them.
permissions.denymust not blockgit worktree …orgit -C …. A blanketgit checkout:*deny is fine — the worker never checks out (the dispatcher usesgit worktree add -b).- If the worker reports
STATUS: isolation-failedon its first verification step and the status file is empty, a missing permission pattern is almost always the cause.
Step 1 — gather context
git rev-parse --abbrev-ref HEAD→$PARENT_BRANCHgit rev-parse --show-toplevel→$PARENT_REPO_ROOT- Detect package manager from lockfile (
pnpm-lock.yaml→ pnpm,yarn.lock→ yarn,package-lock.json→ npm) →$PACKAGE_MANAGER gh repo view --json nameWithOwner -q .nameWithOwner→$GH_REPO(e.g.puzzmo-com/app). Needed so the worker can runghwithout changing directory.- Generate a branch name:
sidequest/<short-slug>(2-4 word kebab-case summary of the task) →$BRANCH_NAME - Compute a slug from
$BRANCH_NAME(replace/with-) →$SLUG $WORKTREE_PATH = /tmp/sidequest-worktrees/$SLUG$STATUS_FILE = /tmp/sidequest-$SLUG.log
Step 2 — create the worktree and status file
mkdir -p /tmp/sidequest-worktreesgit worktree add -b $BRANCH_NAME $WORKTREE_PATH— creates the branch AND the worktree in one step, with the worktree already checked out on the new branch. If this fails (e.g. branch or path already exists), stop and report the error so the user can clean up.: > $STATUS_FILE— create/truncate the status file.
Step 3 — launch the worker
Call the Agent tool with:
subagent_type:"sidequest-worker"run_in_background:true- prompt: only the block below, with variables interpolated:
Task: $ARGUMENTS
Context:
- parent_branch: $PARENT_BRANCH
- parent_repo_root: $PARENT_REPO_ROOT
- branch_name: $BRANCH_NAME
- worktree_path: $WORKTREE_PATH
- package_manager: $PACKAGE_MANAGER
- gh_repo: $GH_REPO
- status_file: $STATUS_FILEDo NOT pass isolation: "worktree". The worktree is managed explicitly in Step 2 — passing the isolation flag would tell the runtime to create a second worktree on the same branch and fail.
Report to the user, in this format:
Sidequest launched on branch `$BRANCH_NAME`.
- worktree: $WORKTREE_PATH
- status: $STATUS_FILE (tail -f to watch progress)
- cleanup: auto-removed on a passing PR; on draft/blocked, run `git worktree remove --force $WORKTREE_PATH` manuallyStep 4 — start the progress loop
Invoke the loop skill via the Skill tool with no interval (self-paced). Pass these args verbatim:
Tail sidequest status file: $STATUS_FILE
On each tick:
1. Read $STATUS_FILE.
2. Compare the lines to what you have already told the user in this conversation. Emit any new lines as one short update each. If nothing is new, stay silent — that is the correct behavior.
3. If a line starting with `DONE` appears, post a final summary and end the loop (do NOT schedule another wake).
4. If a line starting with `STATUS: isolation-failed` or `STATUS: blocked:` appears, post it and end the loop.
5. Otherwise, pace the next tick 120–240s out.
You will also receive an Agent task-completion notification when the worker finishes. When that fires, read $STATUS_FILE one last time, emit any remaining lines, then end the loop.
Keep each update under ~15 words. The user can interrupt at any time.Do not do anything else. The loop skill owns the rest.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

