auto
Skill231 starsChanged 23 days agoApache-2.0View on GitHub
Eager autonomous-completion mode for working through a sequence of tasks (resolving a batch of issues, executing a multi-step plan, clearing a backlog). Activate when the user says to work through issues/tasks "on auto", "autonomously", "eagerly", "one after another", "don't ask", or "keep going without me". Changes the working posture: complete as much as possible without pausing for approval, open pull requests automatically (no permission prompt), and start the next branch/PR immediately instead of waiting for the previous one to merge. Does NOT relax the codebase's quality gates (fmt/lint/tests, verification).
- Commits and pushes
11 other projects use a version of this.
Adopted by
Most recently taken up by
- thesun4sky/jobstacktheir own version, first seen Nov 10, 2026
- raine/skillstheir own version, first seen Oct 21, 2026
- GustavoAlecio/claude-setuptheir own version, first seen Oct 16, 2026
- krixon/skillstheir own version, first seen Oct 11, 2026
- ARYN26/you-are-heretheir own version, first seen Oct 9, 2026
- guardana/guardanatheir own version, first seen Oct 8, 2026
Used with
Projects with it also use plan and review.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- Q00/ouroboros★ 6.4ksame name
- guardana/guardana★ 225same name
- thesun4sky/jobstack★ 216same name
- zjunlp/Mechanist★ 75same name
- assaio/assaio★ 63same name
- raine/skills★ 2same name
- ARYN26/you-are-here★ 1same name
- krixon/skills★ 1same name
- RachaelQuisel/skillsforyou★ 1same name
- Co4an/claude-auto★ 0same name
1 more not listed
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.
In this repositoryOne other file it gives its agents.
- CLAUDE.mdCLAUDE.md
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 25
Work eagerly, end to end: When there is a sequence of tasks in front of you — a batch of open issues, the steps of a plan the user handed …
- Distinctive: line 29
Stop only when the sequence is exhausted, the remaining items are genuinely blocked (see *Where you still stop*), or the user interrupts.
- Very specific: line 52
Keep every quality gate green: Run the full pre-commit checklist before each commit — cargo fmt --all, cargo lint (default features), cargo …
- Very specific: line 58
Follow the codebase rules: Never edit on main; production code must not call .unwrap() (use ?
- Very specific: line 63
Escalate genuine ambiguity: If a task has a load-bearing decision with no clear default, the acceptance criteria are unclear, or a change …
- Distinctive: line 80
Verify: run the quality gates above and exercise the behaviour.
- Very specific: line 81
Commit on a fresh per-task branch (with the session's required commit trailers), push with git push -u origin <branch>.
- Distinctive: line 87
Keep a short running tally so the user can see progress (item → branch → PR URL → status).
“Common” is measured: kept by at least 6 in 10 of the 8 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
auto — eager autonomous completion
auto is a posture modifier, not a standalone workflow. It wraps the
normal task loop (e.g. /resolve-issue, working through a plan, clearing a
list of open issues) and changes how you drive it, not what the underlying
work is. Activating auto is the user's explicit, standing authorization for
the three behaviours below — you do not need to re-ask for them per task.
What changes while auto is active
- Work eagerly, end to end. When there is a sequence of tasks in front of you — a batch of open issues, the steps of a plan the user handed you — carry it forward as far as you can in one go. Do not stop after a single unit and wait to be told "next"; pick the next resolvable item and continue. Stop only when the sequence is exhausted, the remaining items are genuinely blocked (see Where you still stop), or the user interrupts.
- Open the PR automatically — do not ask. The base rule "don't open a PR
unless explicitly asked" and
/resolve-issue's "stop and ask permission to implement" are satisfied by the user activatingauto. Once a change is complete and verified, commit, push, and open the PR without a permission prompt. Report the PR number/URL and keep moving. - Don't wait for merge before the next branch. A PR being open is not a
reason to idle. Immediately branch for the next task and start it. Each task
gets its own fresh branch cut from up-to-date
main(git checkout main && git pull origin main && git checkout -b <descriptive-name>) and its own PR — do not stack unrelated changes onto one branch. If the session pinned you to a single assigned branch, activatingautois the explicit permission to create additional per-task branches; prefer one branch + one PR per task so reviews stay independent.
Where you still stop (auto does not override these)
auto removes approval friction and serialization, not judgement and
quality. You must still:
- Keep every quality gate green. Run the full pre-commit checklist before
each commit —
cargo fmt --all,cargo lint(default features),cargo lint-all(all features — catches feature-gated code), and the relevantcargo testtargets. Actually exercise the change (run the test or example and read the output); a passing typecheck is not verification. Never open a PR on red. - Follow the codebase rules. Never edit on
main; production code must not call.unwrap()(use?/.expect("invariant: …")); add tests in historical mode (RunMode::HistoricalFrom(NanoTime::ZERO)) with hand-verified expected values; update every affected call site (examples, benches, Python bindings) so the whole workspace still builds. - Escalate genuine ambiguity. If a task has a load-bearing decision with
no clear default, the acceptance criteria are unclear, or a change would
restructure the public API/architecture, pause and ask (
AskUserQuestion) rather than guessing — even in auto mode. Eager ≠ reckless. - Skip, don't force, the intractable. An item that needs an unavailable service, a large new dependency stack, or a change you cannot verify here is not "resolvable in one PR". Note that you skipped it and why, then move to the next resolvable item.
Operating loop
For each item in the sequence:
- Select the next resolvable item (highest priority first, per the underlying workflow's selection rule).
- Implement it, mirroring the closest existing node/adapter and the
error-handling rules in
CLAUDE.md. - Verify: run the quality gates above and exercise the behaviour.
- Commit on a fresh per-task branch (with the session's required commit
trailers), push with
git push -u origin <branch>. - Open the PR automatically with a body that says what changed, why, how it
was verified, and a
Closes #<n>keyword when it resolves an issue. - Without waiting for CI or merge, go to step 1 for the next item.
Keep a short running tally so the user can see progress (item → branch → PR URL → status). When the sequence is done, summarise every PR you opened.
De-activating
Auto stays in effect for the rest of the session's batch unless the user says to stop, slow down, or go back to asking per step. When they do, revert to the default cautious posture (ask before opening PRs, one unit at a time).
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.

