agentleFS
ChindanaiNaKubChindanaiNaKub/openthought/.claude/skills/autopilot/SKILL.md

autopilot

Skill0 starsChanged 4 months agoMITView on GitHub

Run an unattended openthought session — pick work from the intent graph with coherence next, build it, gate it, commit it, and keep going until the graph runs dry, escalating questions to the inbox instead of stopping. Use when the user says to keep going, work through the backlog, run on autopilot, ship whatever is next, or leaves you to work without further instructions.

82 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

8 rules. 0 are common, 3 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 73 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    Pick: Run coherence next. Take the top story.

    line 24
  2. Very specific:

    Build: Do the work the story states.

    line 30
  3. Very specific:

    Gate: Run coherence ship. Judge every criterion against the actual diff: met, unclear, or unmet, with the line that proves it.

    line 34
  4. Very specific:

    Drain the queue: When no story is workable, run coherence queue and answer judgment tasks per the policy below.

    line 45
  5. Very specific:

    End: Run coherence brief. Add one line of your own judgment on top: the single most important thing the human should know.

    line 48
  6. Distinctive:

    Never tick a criterion to make a story closeable.

    line 91
  7. Distinctive:

    If you reverted or abandoned work mid-session, the commit history must show it.

    line 95
  8. Distinctive:

    No squashing away the dead end.

    line 96

“Common” is measured: kept by at least 6 in 10 of the 73 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

The human wrote the intent. Your job is to ship it without them, and to leave a trail they can trust in five minutes tomorrow morning.

One principle governs everything here: when in doubt, ask and move on. Never ask and wait. A blocked session that stalls on one question wastes the whole night; a session that queues the question and ships the other four stories wasted nothing.

(Commands assume coherence on PATH; a repo may vendor it at bin/coherence instead. Use whichever exists.)

The loop

  1. Read the inbox. Open intent/_inbox.md. Ticked entries are decisions: apply each one (update the artifact it concerns, then act on it) before doing anything else. The human's answers outrank everything below.
  2. Pick. Run coherence next. Take the top story. If it is draft under an active project, set it active; that is within policy. If next exits 1, go to step 8.
  3. Focus. coherence focus <story-id>. Read the story, its parent project, and everything in its serves: list. If the story contradicts what you learned from a decided inbox entry, ask about it and pick the next story.
  4. Build. Do the work the story states. Stay inside its code: globs; if the work genuinely needs a file outside them, widen the globs in the story as part of the change, and say so in the commit. If the work turns out to need a decision the graph does not contain, ask and move to the next story.
  5. Gate. Run coherence ship. Judge every criterion against the actual diff: met, unclear, or unmet, with the line that proves it. Unclear means the criterion is badly written; ask about it, leave it unticked, and do not mark the story done.
  6. Close. Tick only criteria the diff proves. All ticked: set status: done. Some unticked: leave the story active and record what remains in the story's Notes section.
  7. Commit. Reference the story id in the message. When one file served several stories, name them all in one commit; splitting a coherent change to fake one-commit-per-story is theater. Then go to step 1 — the inbox may matter to the next pick.
  8. Drain the queue. When no story is workable, run coherence queue and answer judgment tasks per the policy below. Then coherence check; fix what is yours to fix.
  9. End. Run coherence brief. Add one line of your own judgment on top: the single most important thing the human should know. Push only if the human has said sessions may push; otherwise leave commits local and say so.

Autonomy policy

The graph is a delegation contract. You operate level 2 and down; levels 0 and 1 belong to the human.

You may, without askingYou must ask, then continue elsewhere
implement, complete, and close active storiescreate, edit, drop, or reparent a bet, goal, the mission, a user, or a problem
activate a draft story under an active projectdrop or reparent any artifact, even a story
create new stories under an active project when the work uncovers themship anything the gate blocks on stale-premise
widen or correct code: globs to match realityresolve a judgment flag that touches a bet, goal, or the mission
answer judgment tasks whose subjects are all stories or projectstick a criterion the diff does not clearly prove
tick criteria the diff proves; mark stories and projects doneact against a ticked inbox answer, however wrong it seems
commitpush, unless standing permission exists

The left column is the job. The right column is the human's sleep being protected: every entry is either a level-0/1 decision or an irreversible one.

Escalating well

An ask entry is read at breakfast by someone with no session context. It needs three things: the decision required, the option you would take, and what you did in the meantime.

coherence ask --about story-export \
  "Criterion 'exports are fast' is not checkable. I read it as under 2s for 10k \
rows, which the diff meets. Left it unticked; story stays active. Confirm the \
threshold or reword the criterion."

Weak escalations kick the work back: "what should I do about story-export?" makes the human reconstruct everything, which is the exact cost this system exists to kill.

Honesty rules

These are not negotiable, because the human is not watching.

  • Never tick a criterion to make a story closeable. An unticked box costs a morning read; a falsely ticked one costs the trust the whole system runs on.
  • Report failures in the brief as plainly as successes. "story-x: attempted, reverted, see inbox" is a good line.
  • If you reverted or abandoned work mid-session, the commit history must show it. No squashing away the dead end; the dead end is information about the graph.

Related: intent-graph, coherence-check, ship-check.

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.