Intent Clarification Skill

Turn vague human intent into language people and agents can act on.

Expression Precision helps people say what they mean, and helps AI agents understand what the user actually wants before they execute.

For peopleFor AI agentsClarify before acting
Intent scan
“Make this page feel more advanced.”
“advanced” hides multiple stylessuccess criteria are missingexecution would be a guess
03
Clarified intent

Do you mean more minimalist, more futuristic, more luxury-brand-like, clearer information hierarchy, or more refined motion?

Two audiences

One skill for people who need to say it, and agents that need to understand it.

For people

Help someone say what they mean to themselves or another person.

It turns low-resolution feelings, feedback, boundaries, and ideas into concrete language without adding false certainty.

“I am annoyed.” → “This is closer to feeling out of control after repeated interruptions.”
For AI agents

Clarify the task before the agent starts doing the wrong thing.

Put it in front of coding, design, writing, support, research, HR, or product agents whenever user intent is underspecified.

“Build something premium.” → “Premium in visual style, information hierarchy, animation quality, or brand tone?”
The difference

Most assistants optimize the sentence. This skill optimizes the understanding first.

01

One-shot rewrite or execution

Makes the wording smoother, or starts executing, while preserving the same ambiguity.

“I can make the page more professional and visually appealing.”
02

Expression Precision

Finds the missing facts, asks targeted choices, and only then produces language or hands off a clarified task.

“Before changing the page, choose which meaning of ‘advanced’ matters: minimalism, technology, luxury, hierarchy, or motion.”
Agentic workflow

Clarify first. Then write, speak, decide, or execute.

The skill keeps the user in control of the meaning. It does not guess motives, invent facts, or let an agent run on an unclear instruction.

01

Detect ambiguity

Locate vague states, emotional labels, big words, unclear references, and instructions that hide multiple possible meanings.

02

Ask high-value choices

Use concise multi-select questions, including a “None of the above” path when the current hypotheses miss the user's meaning.

03

Produce usable intent

Return a precise name, concrete explanation, copy-ready expression, or clarified task that another person or agent can act on.

Where it fits

Anywhere people start with a blurry internal state or an underspecified request.

Personal reflection

Journals, emotion logs, and self-understanding.

Input

“I am really annoyed today.”

Output

This is closer to losing control after being repeatedly interrupted. What you want is not comfort, but to regain your rhythm.

Diary apps, Notion, Obsidian, Apple Journal, AI diary tools

Relationships

Turn “I cannot say it” into something sayable.

Input

“I want to tell a friend I feel uncomfortable, but I do not know how.”

Output

I am not blaming you for not replying. I just suddenly felt that I was expecting more from this relationship than you were.

Partner talks, friendship, family, apologies, boundaries

Workplace

Turn broad judgments into observable feedback.

Input

“This product experience is bad.”

Output

First-time users do not know what to click next. The primary and secondary buttons have similar visual weight, increasing decision cost.

1:1s, product reviews, customer feedback, retros, interviews

Creators

Help writers find the exact emotional texture.

Input

“I want to write a complex kind of loneliness.”

Output

This is closer to invisible loneliness or emotional exile: sitting in a crowd, but feeling as if life has muted you.

Writing, scripts, fiction, brand voice, speeches, lyrics

AI agents

A pre-execution layer for every capable agent.

Input

“Help me make a more advanced page.”

Output

Does advanced mean more minimalist, more futuristic, more luxury-brand-like, clearer hierarchy, or more refined motion?

Coding agents, design agents, support agents, product agents

How to use it

A human-facing Studio, and an intent layer for agent workflows.

Skill

Intent Clarification Skill

Invoke the skill before a coding, design, writing, research, support, HR, or product agent acts on a vague request.

$expression-precision

Help me make this onboarding flow feel more premium.
Studio

Web Studio

A bilingual guided interface that makes the clarification process visible for demos, writing, feedback, and self-expression.

“Make this page feel more advanced.”

Clarified intent
Open the interactive Studio
API & trust

Visitors can try the live agent without pasting an API key.

The hackathon demo uses a deployment-only Qwen key stored in Vercel server-side environment variables. The browser sends user text to the server route; the key never reaches client code.

Server-side project keyPer-IP hourly request limitInput length limitsDeterministic sample mode when the API is unavailable
For production use, deploy your own copy.

A self-hosted copy keeps your model key and budget controls in your own deployment environment.

No setup for judges or visitors

People can open the Studio and immediately test the product. Cost control belongs in rate limits and the model provider dashboard.

BrowserServerModel API
Expression Precision

Clarify the intent before polishing the sentence or running the agent.

Try the guided workflow with your own vague thought, or load the built-in sample when the live API is unavailable.

Launch Expression Precision