Skip to content
Back to skills

Intent First

ASecurity

Clarify user intent before executing tasks. Use when requests are vague, ambiguous, or could be interpreted multiple ways. Helps prevent wasted effort from misunderstanding.

  • 666 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added February 9, 2026
business

Security analysis

A100/100

Scanned February 12, 2026

npx -y skills add majiayu000/claude-skill-registry --skill intent-first --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Intent First?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Intent First
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majiayu000-intent-first/badge)](https://www.skillsdirectory.com/skills/majiayu000-intent-first)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: intent-first
description: Clarify user intent before executing tasks. Use when requests are vague, ambiguous, or could be interpreted multiple ways. Helps prevent wasted effort from misunderstanding.
---

# Intent-First: Clarify Before You Execute

## Purpose

Transform vague requests into clear, actionable specifications before starting work.

## When to Use

- User request is ambiguous or could mean multiple things
- Task scope is unclear
- Success criteria are not defined
- You're about to start significant work

## The Intent Clarification Framework

### Step 1: Identify the Intent Type

| Type | Signal Words | Example |
|------|-------------|---------|
| **Create** | make, build, write, generate | "Make a report" |
| **Modify** | change, update, fix, improve | "Fix this code" |
| **Analyze** | review, check, find, understand | "Review this design" |
| **Decide** | should I, which is better, compare | "Should I use X or Y?" |
| **Learn** | explain, how does, what is | "How does this work?" |

### Step 2: Ask the 5 Clarifying Questions

Not all questions apply to every request. Select the relevant ones:

#### 1. PURPOSE (Why)
> "What problem are you trying to solve?" or "Who will use this and for what?"

#### 2. SCOPE (What)
> "What should be included? What should NOT be included?"

#### 3. FORMAT (How)
> "What format do you expect? (bullet points, prose, code, diagram)"

#### 4. CONSTRAINTS (Limits)
> "Any constraints? (time, length, technology, style)"

#### 5. SUCCESS (Done)
> "How will you know this is successful?"

### Step 3: Confirm Understanding

Before proceeding, summarize:

```
I understand you want to:
- [Action]: [specific task]
- [Purpose]: [why this matters]
- [Output]: [expected deliverable]
- [Constraints]: [limitations to respect]

Shall I proceed?
```

## Examples

### Vague Request
> "Help me with my presentation"

### Clarified Intent
```
I understand you want to:
- Action: Improve/create presentation slides
- Purpose: [Need to ask - what's the presentation for?]
- Output: [Need to ask - slides? script? both?]
- Constraints: [Need to ask - time limit? audience?]

Let me ask:
1. What is the presentation about and who is the audience?
2. Do you have existing slides to improve, or starting from scratch?
3. What do you need help with - content, structure, or visuals?
```

## Anti-Patterns

| Don't Do This | Do This Instead |
|---------------|-----------------|
| Start working immediately on vague requests | Pause and clarify first |
| Ask all 5 questions every time | Select relevant questions only |
| Make assumptions silently | State assumptions and confirm |
| Over-clarify simple requests | Use judgment on when to clarify |

## Quick Decision: When to Skip Clarification

Skip if ALL are true:
- [ ] Request is specific and unambiguous
- [ ] Scope is clearly bounded
- [ ] Failure cost is low (easy to redo)
- [ ] You have high confidence in understanding

## Integration with Other Skills

After clarifying intent:
- Use **delegation-triggers** to decide who should execute
- Use **skill-activator** to find relevant skills for the task

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.