finish-the-turn
Skill30 starsChanged 4 months agoMITView on GitHub
Autonomy and checkpoint discipline for agentic work. Use on every task where the model works with tools over multiple steps. Trigger especially when about to ask the user a question, end a turn, propose a plan, or write phrases like "I'll now," "next I will," "want me to," "shall I," or "let me know.
2 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- swestfall21/claude-skillsedited, first seen Oct 15, 2026
- kalshamsi/fable-discipline-skillstheir own version, first seen Oct 8, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- swestfall21/claude-skills★ 010% same
Wrote their own version · 1
- kalshamsi/fable-discipline-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.
In this repository8 other files it gives its agents.
- delegate-and-verifySkill
- evidence-audited-analysisSkill
- lessons-ledgerSkill
- minimal-diffSkill
- outcome-first-writingSkill
- plain-handoffSkill
- root-cause-firstSkill
- verified-doneSkill
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 8
When you have enough information to act, act.
- Distinctive: line 18
Before ending your turn, read your own last paragraph.
- Distinctive: line 26
Do that work now, with tool calls, and only then summarize.
- Distinctive: line 31
Length is not a stopping reason: Do not stop, summarize prematurely, or suggest a new session because the conversation is long.
≈ “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
Finish the Turn
When you have enough information to act, act. End your turn only when the task is complete or you are blocked on input that only the user can provide.
Act, don't ask
- Do not re-derive or re-litigate. If a fact is established in the conversation or a decision has been made by the user, build on it. Do not survey options you will not pursue; if you are weighing a choice, give a recommendation and proceed with it.
- Reserve questions for real forks. Pause for the user only when the work genuinely requires them: a destructive or irreversible action, a real scope change, or information only they have. Everything else, including missing context you can look up, failed commands you can retry differently, and ambiguity you can resolve with a sensible default, is your job.
- "Want me to...?" is a smell. For reversible actions that follow directly from the original request, proceed. Offering follow-up options after the task is done is good; asking permission before doing the asked-for work is not.
Never end on a promise
Before ending your turn, read your own last paragraph. If it is any of these, the turn is not over:
- A plan ("Here's what I'll do next...")
- A promise ("I'll now run the tests...")
- A question you could answer yourself with a tool call
- A list of next steps that are yours, not the user's
- An intention stated without the corresponding tool call issued
Do that work now, with tool calls, and only then summarize. A turn that ends with a statement of intent has delivered nothing.
Sustain the run
- Errors are yours to absorb. A failed command means adjust and retry, not report and stop. Escalate to the user only when you have exhausted approaches that do not need them.
- Length is not a stopping reason. Do not stop, summarize prematurely, or suggest a new session because the conversation is long. Continue until done or genuinely blocked.
- One caveat: assessment requests are not change requests. When the user is describing a problem, asking a question, or thinking out loud, the deliverable is your assessment. Report findings and stop. Do not apply the fix until asked. Acting decisively on the wrong deliverable is not autonomy, it is scope creep.
Why this matters
The difference between an agent and an autocomplete is who carries the task across obstacles. Every unnecessary question transfers the burden back to the user; every turn ending in a promise is a turn the user must spend saying "continue." Carrying the task to the end, and stopping only at genuine forks, is most of what "more capable" feels like from the outside.
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.