Compatible with Claude, ChatGPT, Gemini, and Claude Code.
Prompt-Master
Turns scattered thoughts into a structured prompt.
Free, no account, no email — the command is all you need.
Already have it?
Nothing to install if it's already active
If this skill is already added in your Claude, there's no need to install it again. Just ask for what you want to do.
Not sure? Ask Claude: "which skills do you have available?"
“Use the prompt-master skill to turn this idea into a structured prompt ready to paste into any AI tool: …”
1 · The simplest way
Copy and paste into any AI chat
This button copies the whole skill. Paste it into Claude, ChatGPT, or Gemini, then describe what you need. No installation required.
In a chat, Word or Excel files can't always be generated: you'll receive the document as structured text instead, and Claude will flag it.
2 · Command line
/prompt-master installs in 3 steps
For Claude Code. The skill installs on your computer and stays available across all your projects.
Prerequisites: Claude Code + Node.js
The skill runs inside Claude Code, Anthropic's AI assistant installed on your computer. Node.js is the engine that runs the install command. Both are free.
Open PowerShell
Click Start, type "PowerShell", then press Enter.
Paste the command
Right-click inside the PowerShell window (or press Ctrl + V) to paste the command below, then press Enter. Wait a few seconds.
npx educasium --skill prompt-master --globalWhat you should see:
npx educasium --skill prompt-master --global
Fetching https://raw.githubusercontent.com/Chappygo-OS/claude-meta-skills/main/dist/prompt-master.zip
…
Installed "prompt-master". Restart Claude Code to pick it up.
Restart Claude Code
In Claude Code, type /restart (or quit and relaunch Claude Code). The skill now appears in the command list.
Don't use a terminal?
Using Claude.ai in your browser?
The assisted install works through a .zip file imported into Claude.ai. It takes longer and asks for your email.
What the skill does
View the source codeIn detail.
Describe what you want, even loosely. Prompt-Master asks the missing framing questions (role, audience, constraints, format) and rebuilds a prompt ready to paste into any AI tool.
When to use it
Before any important AI task. The time saved on output far outweighs the framing time.
Example usage
/prompt-master I need a follow-up email for a client...In depth
How this skill really works
Prompt-Master turns a scattered thought — especially a voice-transcribed input — into a structured, optimized prompt ready to paste into the target AI tool (Claude, ChatGPT, o3, Gemini, Midjourney, Cursor, Claude Code, ComfyUI, and about twenty other categories documented in the skill). It always starts by identifying the target tool, then extracts 9 intent dimensions (task, target tool, output format, constraints, input, context, audience, success criteria, examples), then runs a diagnostic against 35 credit-killing patterns (a vague task verb, two tasks crammed into one prompt, no success criteria, etc.), documented in references/patterns.md.
The skill applies non-negotiable hard rules: never add Chain of Thought to a reasoning-native model (o3, o4-mini, DeepSeek-R1, Qwen3 thinking mode) — it degrades output —, never embed techniques that cause fabrication in single-prompt execution (Mixture of Experts, Tree of Thought, Graph of Thought, Universal Self-Consistency, all of which require real external routing or sampling), and never ask more than 3 framing questions.
Output is always a single copyable prompt block, followed by a line naming the target tool and what was optimized — never an unrequested explanation. For complex multi-step tasks, the skill can produce a numbered prompt chain where each step consumes the previous step's output.
When to use it
- Before any important AI task, starting from a scattered thought or a voice-transcribed input.
- When the user doesn't yet know which target tool to use and the category needs identifying first (LLM, agentic IDE, image, video, no-code workflow...).
- When the user pastes an already-written prompt that's underperforming — to diagnose which of the 35 patterns explains the problem.
- When the task needs several sequential steps, each consuming the previous step's output.
- In auto-mode, added to CLAUDE.md to screen every long, unstructured prompt before processing.
- When the prompt targets a reasoning-native model (o3, o4-mini, DeepSeek-R1, Qwen3 thinking) and adding Chain of Thought by mistake needs to be avoided.
When not to use it
- Not for a short, clear, one-sentence request ("translate this to French") — optimization adds nothing.
- Not when the user already wrote a well-structured prompt — don't rework it without reason.
- Not when the scattered thought actually describes building an entire workflow or skill — use /process-interviewer for that instead.
- Not for cleaning AI-writing tells out of a final text — that's /humanizer, to run on the optimized prompt itself if needed.
The method
The 9 intent dimensions
Before writing any prompt, the skill silently extracts 9 dimensions documented in the SKILL.md. Dimensions marked critical trigger a clarifying question when missing — capped at 3 questions total.
- Task
- The precise action requested — convert any vague verb into an exact operation. Always critical.
- Target tool
- Which AI system receives this prompt — determines the syntax and template to apply. Always critical.
- Output format
- The shape, length, structure, and filetype expected for the result. Always critical.
- Constraints
- What MUST and MUST NOT happen. Critical if the task is complex.
- Input
- What the user is providing alongside the prompt. Critical if applicable.
- Context
- Domain, project state, prior decisions. Critical if the session has history.
- Audience
- Who reads the output and their technical level. Critical if the output is user-facing.
- Success criteria
- How to know the prompt worked — binary where possible. Critical if the task is complex.
- Examples
- Desired input/output pairs to lock a pattern. Critical if format is make-or-break.
Steps
Identify the target tool
Route to the correct template and syntax among the roughly twenty documented categories (Claude, GPT, o3, Gemini, Midjourney, Cursor, Claude Code, ComfyUI...). If unclear, ask once.
Extract intent across 9 dimensions
Silently extract the 9 dimensions; only ask a clarifying question for ones marked critical and missing, never more than 3 total.
Detect task type
Classify the task (creative writing, research/analysis, code generation, decision support, data processing, communication, ideation) to apply the right optimization strategy and the right template among the 17 documented.
Apply the 35-pattern diagnostic
Scan the raw prompt for the documented failure patterns (task, context, format, scope, reasoning, agentic) and fix them silently — flag only if the fix changes intent.
Apply safe techniques
Only apply role assignment, few-shot examples, grounding anchors, or Chain of Thought when the task genuinely needs them — never Chain of Thought on a reasoning-native model.
Produce the output
Return a single copyable prompt block, with the target tool and one sentence on what was optimized. For complex tasks, produce a numbered prompt chain where each step consumes the previous output.
Rules
- Never return a prompt without first confirming the target tool — ask if ambiguous.
- Never embed techniques that cause fabrication in single-prompt execution (Mixture of Experts, Tree of Thought, Graph of Thought, Universal Self-Consistency).
- Never add Chain of Thought to a reasoning-native model (o3, o4-mini, DeepSeek-R1, Qwen3 thinking).
- Never ask more than 3 clarifying questions before producing a prompt.
- Never add an explanation the user didn't ask for — output stays a copyable prompt block plus one "tool + why" line.
- Always verify the reworded prompt keeps the original intent before silently fixing one of the 35 patterns.
Mistakes to avoid
- Adding "think step by step" to a prompt meant for o3 or DeepSeek-R1 — actively degrades their output.
- Leaving an agentic prompt (Claude Code, Cursor) with no stop condition or explicit file scope.
- Merging two distinct tasks into one prompt instead of delivering them as Prompt 1 / Prompt 2.
- Using Mixture-of-Experts or Tree-of-Thought-style techniques in single-pass execution, promising routing or branching that doesn't actually exist.
- Delivering a Midjourney prompt written as full prose instead of comma-separated descriptors.
- Asking the user more than 3 clarifying questions before producing a first version.
What this skill doesn't do
- Does not execute the task described in the prompt — it only produces the prompt to paste into the target tool.
- Does not verify the factual accuracy of claims inside the prompt — that's /fact-checker.
- Does not build a full workflow or skill from a scattered thought — that's /process-interviewer.
- Does not clean AI-writing tells out of the final output — that's /humanizer, to run on the optimized prompt if needed.
- Does not actually invoke the recommended model (Haiku/Sonnet/Opus) — the routing signals are conventions added as a comment, not an execution.
Skill contents — 5 fichiers · 46.7 KB(SKILL.md, references/chain-patterns.md, references/patterns.md, references/templates.md, references/tool-routes.md)
Download for free