intent-modeling — Claude Code skill: read the goal, not the words
A skill that makes an AI agent work out what you meant before it acts on what you literally said.
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a claude-codeWhat intent-modeling does
The skill pushes the agent to read for the outcome you want instead of the words you used. Conversation context and earlier turns feed that inference. When the goal stays ambiguous, the skill says to ask rather than guess quietly.
Corrections get their own treatment. A note like less of X, more of Y usually calibrates a balance, not a permanent ban, so the skill tells the agent to carry that balance into the next step. Depth follows the request too: brief when you want brevity, thorough when you want depth. A vague ask may just be a rough draft of the real need.
Here is the part that earns the install. A correction names one instance of a mistake, and the mistake is usually a class. Before replying, the agent hunts for the other instances in the work and the system around it, fixes the ones inside its task and role, reports the rest with a proposed follow-up, and says where it looked. Runs on Claude, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode.
What it helps with
- work out what a person meant, not just what they typed
- keep a directional correction as a balance instead of a ban
- match answer length to the request
- find the other spots with the same mistake
- ask a question when the task is vague
Example prompts for the agent
- Work out what I actually want here and do that
- I asked for shorter — check where else the text has the same problem
- Before you edit, tell me how you read the task
How to install intent-modeling
1Skills CLI (Vercel)
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a claude-code2Manually: put the files in place
~/.claude/skills/intent-modeling/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/intent-modeling/
1Skills CLI (Vercel)
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a codex2Manually: put the files in place
.agents/skills/intent-modeling/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a cursor2Manually: put the files in place
.cursor/skills/intent-modeling/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install haowjy/creative-writing-skills intent-modeling2Manually: put the files in place
.github/skills/intent-modeling/SKILL.md1Skills CLI (Vercel)
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a gemini-cli2Platform CLI
gemini skills install https://github.com/haowjy/creative-writing-skills --path cw/skills/intent-modelingManually: put the files in place
.devin/skills/intent-modeling/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add haowjy/creative-writing-skills --skill intent-modeling -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about intent-modeling
How do I install intent-modeling in Claude?
Run npx skills add haowjy/creative-writing-skills --skill intent-modeling -a claude-code. The skill file lands at ~/.claude/skills/intent-modeling/SKILL.md.
What does intent-modeling do?
It helps the agent separate what you said from what you meant: it reads conversation context, infers the underlying goal, and asks when that goal is unclear. Corrections are treated as balance calibration rather than a ban.
Which platforms does the skill support?
Claude, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode. Each platform has its own install path in the repository.
Source text
SKILL.md · Apache-2.0
---
name: intent-modeling
description: 'Use before acting on human instructions: separate what they said from what they meant.'
---
# Intent Modeling
Read for the outcome the human wants, not only the words they used. Use
conversation context and prior turns to infer underlying goal. When the
goal is ambiguous, ask.
Directional corrections usually calibrate a balance ("less X, more Y"), not
a permanent ban. Encode the target balance in what you do next.
Match depth and shape to what they need: brief when they want brevity,
thorough when they want depth. A vague request may be a rough draft of
the real need. Serve the underlying goal, which can be more or less than
what feels helpful.
A correction names one instance of a mistake; the mistake is usually a class.
Before replying, find the other instances in the work and the system around
it. Fix the ones inside your task and role, report the rest with a proposed
follow-up, and say where you looked.
© haowjy · Apache-2.0 · haowjy/creative-writing-skills
