clarify-first
Skill0 starsChanged 7 days agoMITView on GitHub
Prevents misunderstanding before work starts. Use when a request is ambiguous, underspecified, large, or multi-step (for example "build a dashboard", "write the report", "refactor this module"), or when the user's goal, audience, scope or definition of done is unclear. Asks at most three targeted questions with defaults, then restates the goal and done criteria. Skip it when the request is already clear.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- OrionicsLtd/claude-skillstheir own version, first seen Nov 1, 2026
- M1688-cpu/ask-firsttheir own version, first seen Oct 6, 2026
- geniusBrainLsm/clarify-firsttheir own version, first seen Aug 13, 2026
- DmiyDing/clarify-firsttheir own version, first seen Mar 12, 2026
- FaisalAlqarni/sp-ecctheir own version, first seen Feb 6, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- FaisalAlqarni/sp-ecc★ 9same name
- DmiyDing/clarify-first★ 6same name
- geniusBrainLsm/clarify-first★ 4same name
- M1688-cpu/ask-first★ 1same name
- OrionicsLtd/claude-skills★ 1same 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 repository11 other files it gives its agents.
- drop-in/AGENTS.mdAGENTS.md
- llms.txtllms.txt
- coverage-checkSkill
- fresh-eyes-reviewSkill
- fresh-start-handoffSkill
- grounded-answerSkill
- honest-critiqueSkill
- learning-tutorSkill
- prompt-improverSkill
- task-specSkill
- verify-claimsSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 6 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
6
Distinctive
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 13
Decide whether to ask: Ask only if a wrong guess would change the result in a way that matters (goal, audience, scope, format, constraints, …
- Distinctive: line 22
Not doing: anything you are deliberately leaving out
- Distinctive: line 23
Surface assumptions: If you proceed on defaults, list the assumptions at the top of your answer and mark with ⚠️ any that would change the …
- Distinctive: line 27
When requirements have piled up over many turns and the output keeps missing, stop patching.
- Distinctive: line 31
Don't ask about things you can find out yourself by reading the files, docs or data you have access to.
- Distinctive: line 32
Don't ask more than three questions, and don't ask when the answer wouldn't change what you do.
“Common” is measured: kept by at least 6 in 10 of the 5 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
Clarify first
Models rarely notice ambiguity on their own, and they tend to lock in early assumptions. A few targeted questions before starting are cheaper than redoing the work.
Steps
- Decide whether to ask. Ask only if a wrong guess would change the result in a way that matters (goal, audience, scope, format, constraints, data source). If the request is clear, skip to step 3.
- Ask at most three questions, the ones with the biggest impact on the result. For each, give a default:
- Who will read this? (Default: non-technical managers.)
- ... If you don't answer, I'll go with the defaults.
- Restate the task in two or three lines before doing it:
- Goal: what the result is for
- Deliverable: what you will produce, in what format
- Done when: how the user will judge it finished
- Not doing: anything you are deliberately leaving out
- Surface assumptions. If you proceed on defaults, list the assumptions at the top of your answer and mark with ⚠️ any that would change the result significantly if wrong.
If the conversation has drifted
When requirements have piled up over many turns and the output keeps missing, stop patching. Consolidate every requirement and decision so far into one prompt and suggest the user start a fresh chat with it.
Don't
- Don't ask about things you can find out yourself by reading the files, docs or data you have access to.
- Don't ask more than three questions, and don't ask when the answer wouldn't change what you do.
Evidence
- Selective clarification (asking only on the most uncertain cases) doubled the gain compared with asking at random: Zhang & Choi, 2023 — https://arxiv.org/abs/2311.09469
- Models are poor at noticing ambiguity unaided: CLAMBER, 2024 — https://arxiv.org/abs/2405.12063
- Spreading requirements over several turns cut performance by 39% on average; consolidating and restarting helps: Laban et al., ICLR 2026 — https://arxiv.org/abs/2505.06120
- More in the guide: https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/04-failure-modes.md#1-misunderstanding
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.

