π Stop prompting like it's 2023! A systematic framework and agent skill for engineering sharp, reliable AI prompts.
Turn vague one-liners into structured, context-rich prompts that get the best out of modern LLMs. Drop it in, say "make this a prompt," and watch the magic happen πͺπ₯
Most prompt failures come from three missing ingredients: unclear expertise, vague objectives, and absent constraints. The ART Framework fixes all three with a simple tripartite structure applied to every prompt it writes or refactors:
| Component | Focus | What it does |
|---|---|---|
| π A: Act As | Persona & Expertise | Sets the AI's perspective, seniority, and domain authority. Replaces generic "expert" labels with a specific professional identity. |
| π― R: Request | Mission & Context | Defines the deliverable, the inputs, the audience, and what is out of scope. |
| βοΈ T: Terms | Rules & Guardrails | Sets the boundaries: format, length, tone, sources, and things to avoid. |
- π― Fewer invented facts: Explicit terms and negative constraints keep the output anchored to what you actually gave it.
- β‘ Copy-paste ready: Every output is a self-contained block that works in a brand-new chat.
- π§© Works beyond Claude: The skill is a single
SKILL.mdfile. The prompts it writes are plain text, so they run in any LLM. - π¬ Repeatable results: Replaces trial-and-error prompting with a structure you can reuse and improve.
Here is how a one-line request becomes an actionable ART prompt:
"Help me write a cold email to a SaaS founder."
ACT AS:
A seasoned B2B sales copywriter who writes short, specific cold emails
that busy SaaS founders actually read.
REQUEST:
Write one cold email to the founder of a 20-person B2B SaaS company that
sells inventory software to small retailers. Goal: book a 15-minute call.
Use only the company name and the founder's first name as placeholders.
TERMS:
1. Length: 75 to 100 words.
2. Tone: direct, peer-to-peer. No buzzwords, no exclamation marks,
no "I hope this finds you well."
3. Structure:
- Line 1: one specific observation about the company (no generic praise).
- Line 2: the value, with one concrete benefit.
- Line 3: one low-friction call to action.
4. Do not invent facts about the company. If something is unknown, leave it out.Clone the repository into your personal skills directory:
git clone https://github.com/YourAKShaw/art-prompt-framework.git ~/.claude/skills/art-prompt-frameworknpx skills add YourAKShaw/art-prompt-framework- Download this repository as a ZIP archive.
- Make sure
SKILL.mdis at the root of the extracted folder. - Upload the folder or ZIP via Claude Settings β Capabilities / Skills.
Once installed, ask your agent to write, improve, or convert a prompt, or mention the ART framework by name:
"Write an ART prompt to help me build an automated financial reconciliation pipeline in Python."
"Turn this into a strong prompt using the ART framework: 'Create a system design document for a real-time chat application.'"
The skill returns a clean, self-contained prompt in a code block, ready to paste into a new chat.
| Environment | Status |
|---|---|
| π οΈ Claude Code | β Supported |
| π Claude.ai | β Supported |
| π¦ Other agents (via Skills CLI) | π§ͺ Check your agent's docs for where skills are loaded |
art-prompt-framework/
βββ π SKILL.md # The skill: framework, conversion rules, output rules
βββ βοΈ .github/ # Commit message workflow and checker
βββ π README.md # Project documentation
βββ π SECURITY.md # Vulnerability reporting policy
βββ βοΈ LICENSE # MIT License
βββ π‘οΈ .gitignore # Git ignore rules
Contributions, feedback, and new prompt patterns are warmly welcomed! π
- Fork the project.
- Create your feature branch (
git checkout -b feature/new-prompt-pattern). - Commit your changes (
git commit -m 'feat: add new terms template'). - Push to the branch (
git push origin feature/new-prompt-pattern). - Open a Pull Request.
For a large change to the framework itself, open an issue first so we can agree on the approach.
Found a vulnerability? Please report it privately as described in SECURITY.md instead of opening a public issue.
Distributed under the MIT License. See LICENSE for full details.