agentleFS
huntiehuntie/skills/autopilot/SKILL.md

autopilot

Skill0 starsChanged 21 days agoMITView on GitHub

Turn on autopilot for the rest of the session. The strategy is to continuously commit, when settled. Each unit of work is committed the moment it is proven to work, without asking, so a task builds a stack of atomic commits on its branch as it proceeds. Takes a rung (commit, push, submit) and `squash`, which tidies each task's stack at the end of the task. Use only when the user invokes /autopilot; `/autopilot off` turns it off.

74 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use review-bugbot, rename-chat and 10 more.

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.
  • 0mlkrizzz655335vSkill

    review-bugbot

    缺陷导向代码审查。用 Bugbot 子代理找 bug。

    Also in 3 projects that use this

  • 0mlkrizzz655335vSkill

    rename-chat

    重命名当前聊天。按聊天内容改标题。"disable-model-invocation: true

    Also in 3 projects that use this

  • Tsumugii24Skill

    split-to-prs

    Split current work into small reviewable PRs. Use when the user asks to split a chat, set of changes, branch, or PR.

    Also in 4 projects that use this

  • PandelisZSkill

    update-cli-config

    View and modify Cursor CLI configuration settings in ~/.cursor/cli-config.json. Use when the user wants to change CLI settings, configure permissions, switch approval mode, enable vim mode, toggle display options, configure sandbox, or manage any CLI preferences.

    Also in 3 projects that use this

  • 32103210Skill

    migrate-to-skills

    Convert 'Applied intelligently' Cursor rules (.cursor/rules/*.mdc) and slash commands (.cursor/commands/*.md) to Agent Skills format (.cursor/skills/). Use when you want to migrate rules or commands to skills, convert .mdc rules to SKILL.md format, or consolidate commands into the skills directory.

    Also in 3 projects that use this

  • 32103210Skill

    update-cursor-settings

    Modify Cursor/VSCode user settings in settings.json. Use when you want to change editor settings, preferences, configuration, themes, font size, tab size, format on save, auto save, keybindings, or any settings.json values.

    Also in 3 projects that use this

  • +1 more from 32103210
  • +1 more from 0mlkrizzz655335v

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 66 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Only commit is defined below.

    line 24
  2. Distinctive:

    Commit the moment a unit settles, and before the next one starts.

    line 32
  3. Distinctive:

    Write every message with the commit-msg skill.

    line 43

“Common” is measured: kept by at least 6 in 10 of the 66 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Autopilot

Continuously commit, when settled. A commit lands at every unit boundary, once that unit is proven to work, so the branch is a stack of atomic commits by the end of the task: each holds one concern, stands on its own, and the stack reads as an argument for the change: every commit ends in a state a reviewer can check, in an order that makes the case.

Takes a rung, squash, or off, and each invocation restates the mode in full: /autopilot push squash sets both, /autopilot commit afterward drops squash. Invoking it is the explicit ask the standing rule against unasked commits requires, and the mode stays on until off or the end of the session. Say "Autopilot is on" in one line when it starts, naming the rung and squash when set; a conversation summary carries that line so the mode survives it.

Rungs

Each rung includes the ones below it. The argument names how far the agent carries work without asking; everything above the current rung remains an explicit ask.

RungFinished without askingReversibility
commit (default)units on the task branchlocal, fully reversible
pushthe branch and its stack layers on originvisible, reversible
submitpull requests, draft firstreviewers are notified

Only commit is defined below; push and submit get sections when a task first needs them.

Settled

A unit is settled when its check has run and passed: the tests for what changed, and lint and typecheck where they apply. Green is the bar. A deliberately red commit, such as the failing test ahead of its fix, is allowed only when that red-then-green order is the delivery story.

Continuously

Commit the moment a unit settles, and before the next one starts. The boundaries:

  • A concern is complete and checked.
  • The next edit belongs to a different concern. Finish and commit this one first.
  • A prep step stands on its own: a rename, a deletion, a scaffold. It goes in its own commit ahead of the change that needs it.

Stage by path, never the whole tree. A commit holds one unit's files; changes already in the working tree when the mode started, and anything unrelated the user parked there, stay out.

How to commit

  • Before the first commit in a checkout, confirm the branch. Work in a checkout lands on a task branch per AGENTS.md, never on the default branch; if the checkout is still on its default branch, start the branch first with scripts/pull and scripts/checkout <owner/repo>@<name> --new. This repo, vault/, and dotfiles/ commit on main. A Sapling checkout of a shared repo takes no named branch: commit drafts with sl on top of the synced default branch.
  • Write every message with the commit-msg skill; it also carries the git and Sapling commands.
  • Amend only the latest commit, only for a correction to it (a missed file, a lint fix, a typo), and only while it is unpushed. Anything else is a new commit. Reordering and squashing the stack are the user's call afterward, unless squash is on.
  • When a gh stack is checked out, commit on the layer that owns the concern per the gh-stack skill. The commit rung produces commits, not layers. In a Sapling checkout, per AGENTS.md, each commit is a layer: commit with sl on the task branch, or on the draft stack itself in a shared repo.

Squash

A task that iterates leaves follow-ups in its stack: the spacing tuned twice, the label renamed, the lint fix. With squash on, the stack is squashed once the task is done: fix-ups fold into the commits they fix, distinct changes stay distinct, and the result may be one commit or several. Prove the tip with the repo's checks before history moves, rebuild on a detached head, and confirm the tree matches before pointing the branch at it; pushed commits are never folded. Off by default, since it rewrites history the user has not seen.

  • What: The unpushed commits since the tip the task started from. Record that tip when the task begins. A follow-up to a commit from before it stays its own commit. A task that touched several checkouts tidies each.
  • When: At the end of the task, before the closing recap, and ahead of anything the push rung pushes. Never at unit boundaries: units land as they settle, so a red check mid-task leaves a stack to inspect.
  • Gate: The checks decide. Red leaves the stack as it is, and the recap says so. A fold that is not a clear call is not made.

Reporting

Name each commit as it lands, hash and title on one line, and list them all in the closing recap. With squash on, the recap lists the stack as it stands instead, hash and title per commit with the commits each absorbed. On off, stop committing, confirm the mode is off, and say what is left uncommitted in each checkout.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.