agentleFS
irfanannaafi19irfanannaafi19/Clarify-Intent/clarify-intent/SKILL.md

clarify-intent

Skill1 starsChanged 3 months agoMITView on GitHub

Use this skill whenever the user's instruction is too generic or unspecific to act on directly, whether or not a file is attached. Ambiguous instructions that MUST trigger this skill include "analyze this file", "fix this document", "please process this", "make a report from this data", "summarize this". This skill asks the user what they actually mean (direction, purpose, desired output) before execution, so Claude does not guess and risk hallucinating. Do not use it for instructions that are already specific and clear about their target, e.g. "extract all invoice numbers from this PDF into an Excel table".

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. Hugo-DDT/clarify-intent★ 26same name
    2. daniel-kindl/dkstack★ 1same name
    3. ZadAnthony/meta-workflow★ 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.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    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
  1. Distinctive:

    If the user's instruction is ambiguous (with or without an attached file), do NOT start the task or make assumptions.

    line 27
  2. Distinctive:

    If a file is attached, it's fine to skim it (title, table of contents, general structure) to help write relevant questions — but don't …

    line 28
  3. Distinctive:

    After the 5-7 questions above, ALWAYS add 1 closing question at the very end: whether the user has any additional information not covered …

    line 40
  4. Distinctive:

    If the user doesn't answer all the questions or only answers some, make a reasonable assumption for the unanswered ones, briefly state that …

    line 44
  5. Distinctive:

    Don't ask something whose answer can already be inferred from the conversation context or the file's content.

    line 48
  6. Distinctive:

    Don't run more than one round of questions.

    line 49

≈ “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.

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

Clarify Intent

A skill for detecting ambiguous user instructions and asking for clarification before starting the task, so the work matches what the user actually needs instead of Claude's own assumptions.

When an instruction counts as ambiguous

An instruction is considered ambiguous if any of the following apply:

  • It's too generic and could have multiple valid interpretations (e.g. "analyze this file", "fix this document", "make a report", "process this data")
  • It's unclear what output is expected (a document? a summary? a table? bullet points?)
  • It's unclear what focus or angle is being asked for (e.g. the file covers many topics and it's unclear which one is meant)
  • A file is attached but the instruction doesn't say which part to use or for what purpose

An instruction is considered clear and does NOT trigger this skill when it already specifies a target, format, and scope, for example:

  • "Extract all invoice numbers from this PDF into a table"
  • "Translate the 3rd paragraph on page 5 into English"
  • "Make a bar chart from the sales column in this file"

How it works

  1. If the user's instruction is ambiguous (with or without an attached file), do NOT start the task or make assumptions.
  2. If a file is attached, it's fine to skim it (title, table of contents, general structure) to help write relevant questions — but don't start working on the actual task content yet.
  3. Compose 5-7 short clarifying questions covering:

    • The main purpose/intent of the request
    • Direction or focus of the work (which part, which topic, which angle)
    • Desired output (format: document, table, summary, bullet points, etc.)
    • Scope (whole file/specific section, all topics/specific topic)
    • Desired length or level of detail (if relevant)
    • Audience or context of use (if relevant)
    • Other constraints worth knowing (tone, things to avoid, etc.)

    The number of questions MUST always be 5-7, never fewer, even when the available context is minimal (e.g. no file or text at all). If the categories above aren't enough to reach 5-7 questions, add other relevant questions, such as: status of the material/file (attached or not), the intended data source, expected type of fix/change, what may or may not be changed, priority if there are many parts, and other technical preferences relevant to the task. The goal is for a single round of questions to gather everything needed, including the case where the material isn't available yet.

    After the 5-7 questions above, ALWAYS add 1 closing question at the very end: whether the user has any additional information not covered by the previous questions (open-ended answer, not multiple choice). This makes the total 5-7 core questions + 1 closing question.

  4. Ask the questions using an interactive choice tool if available (e.g. ask_user_input_v0), with short, clear answer options for the core questions. For the closing question, ask it as an open-ended question (not multiple choice) since it's meant to capture free-form input from the user. If no such tool is available, ask all questions (including the closing one) as a short numbered list in plain text.
  5. Once the user answers, EXECUTE the task immediately based on those answers. Don't re-summarize your understanding or ask for additional confirmation — the answers to the clarifying questions are treated as the final instruction.
  6. If the user doesn't answer all the questions or only answers some, make a reasonable assumption for the unanswered ones, briefly state that assumption, and proceed with execution anyway.

Key principles

  • Don't ask something whose answer can already be inferred from the conversation context or the file's content.
  • Don't run more than one round of questions. Once the user answers (fully or partially), proceed — don't ask again in a second round.
  • Questions must be short, specific, and easy to answer (ideally answerable with a short choice, not an essay).
  • This skill doesn't apply to instructions that are already specific — don't slow down simple tasks with unnecessary questions.

Examples

Ambiguous (this skill activates):

  • "analyze this file" (with a financial report PDF attached)
  • "fix this document"
  • "please process this file"
  • "make a report from this data"
  • "summarize this"

Clear (this skill does NOT activate, execute directly):

  • "extract all invoice numbers from this PDF into an Excel table"
  • "fix typos and grammar in this document, don't change the content"
  • "make a 1-page executive summary of this report for the board"

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.