heads-down
Skill1 starsChanged 2 months agoMITView on GitHub
Run all the way to the goal with no check-ins. Use when the user invokes /heads-down or says to just go, run to the end, don't stop to ask. Do not stop for steps already in the plan or ask "go?"; stop only for what only the user can give.
In this repository39 other files it gives its agents.
- apple-native-motionSkill
- apple-ui-copySkill
- audit-macos-appSkill
- brain-wikiSkill
- catch-me-upSkill
- checkpointSkill
- claude-md-coachSkill
- claudeSkill
- codexSkill
- crash-debuggingSkill
- cut-memorySkill
- explaining-problemsSkill
- file-batch-safetySkill
- handoffSkill
- hopper-mcpSkill
- icon-designSkill
- interface-integritySkill
- ios-defaultsSkill
- kitSkill
- macos-defaultsSkill
- mock-and-pickSkill
- obsess-heroSkill
- obsessSkill
- offline-archiveSkill
- os-workaroundsSkill
- paper-mcpSkill
- predictionSkill
- second-loopSkill
- session-auditSkill
- ship-macosSkill
- show-meSkill
- slice-and-reviewSkill
- slow-modeSkill
- spatial-interfacesSkill
- standupSkill
- synthetic-fixturesSkill
- transitional-interfacesSkill
- web-design-defaultsSkill
- window-captureSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 33
When a stop is needed, do everything else first, then raise it once at the end, with the options.
- Distinctive: line 38
Commit as you go, one coherent change per commit, explicit paths only.
- Distinctive: line 41
Stop processes you started and clean up scratch files before reporting.
- Distinctive: line 50
Needs you: only the items from "Stop only for", if any.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Heads down
/heads-down means run all the way to the goal. The user has already said yes to the plan. Stopping to ask for approval defeats the point.
This changes how often we talk, not what you're allowed to do. Permissions, verification, and safety rules stay the same.
Start
From the request and the repo, get clear on: the outcome, what's in scope, and what proves it's done. If two readings would produce really different work, ask one tight question. Otherwise start right away. Don't restate the plan back.
Don't stop for
- Steps already in the plan or obviously implied by it.
- "Go?", "Should I continue?", "Want me to do the next part?" The answer is yes.
- Choices with a convention or an obvious default. Make the call and note it for the report.
- A failed approach. Try a genuinely different one.
- Small fixes you find along the way that serve the goal.
Stop only for
- Credentials, auth, or anything security shaped.
- Spending money or signing up for anything.
- Anything leaving the machine: pushing to a public remote, opening a PR, posting, email.
- Deleting outside quarantine, or a batch operation whose full effects haven't been seen.
- Real ambiguity about what to build, where a wrong guess wastes the run or changes the product.
- A wall: two genuinely different approaches both failed. Report the wall and the evidence. Don't force it with disabled checks, history rewrites, or destructive resets.
When a stop is needed, do everything else first, then raise it once at the end, with the options.
While working
- Batch independent reads and checks. Use subagents when they help; pick their model by task weight.
- Commit as you go, one coherent change per commit, explicit paths only.
- New messages from the user are steering. Re-aim right away.
- Park out-of-scope ideas for the report instead of building them.
- Stop processes you started and clean up scratch files before reporting.
Report
First line: what landed and how to try it, or what didn't land and why.
- Verified: the checks you ran and what you saw.
- Calls: non-obvious decisions you made. Empty is fine.
- Parked: useful things outside scope. Empty is fine.
- Needs you: only the items from "Stop only for", if any.
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.

