prompt-factory
Skill17 starsChanged 4 months agoMITView on GitHub
Generate world-class, production-ready prompts in one shot through intelligent 5-7 question flow and 69 presets across 15 professional domains. Supports XML/Claude/ChatGPT/Gemini formats with quality validation. Returns a single mega-prompt ready to use—skill does NOT implement the work described in the prompt. Optimized for modern LLMs including Claude 3.5+, GPT-4.1, and Gemini 2.5.
2 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- Lucien-1127/prompt-factorytheir own version, first seen Oct 23, 2026
- nilecui/SkillsBaseedited, first seen Dec 27, 2025
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- nilecui/SkillsBase★ 2711% same
Wrote their own version · 1
- Lucien-1127/prompt-factory★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository60 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-browserSkill
- agent-engineeringSkill
- app-marketing-copySkill
- ascii-pixel-artSkill
- autopredictSkill
- autoresearchSkill
- babysit-prSkill
- bird-fastSkill
- blockchain-auditorSkill
- brainstormingSkill
- ceo-reviewSkill
- ck-reviewSkill
- cloudflare-managerSkill
- codex-advisorSkill
- content-trend-researcherSkill
- continuous-learningSkill
- coreml-optimizerSkill
- cto-reviewSkill
- direct-mail-strategistSkill
- evangelion-designSkill
- executing-plansSkill
- freshdesk-apiSkill
- gemini-image-generatorSkill
- gemini-visualSkill
- github-actions-templatesSkill
- googleSkill
- gsplat-optimizerSkill
- hyperagentSkill
- idea-mdSkill
- imagegenSkill
- ios-app-testerSkill
- llm-advisorSkill
- llm-evaluationSkill
- macos-appsSkill
- markdown-fetchSkill
- mcp-builderSkill
- mcp-testerSkill
- nano-bananaSkill
- opensea-apiSkill
- paperclipSkill
- playdate-devSkill
- poke-assistantSkill
- poplar-direct-mailSkill
- practical-typographySkill
- qmdSkill
- scientific-critical-thinkingSkill
- security-best-practicesSkill
- security-threat-modelSkill
- seo-optimizerSkill
- shadertoySkill
- skill-finderSkill
- skimmableSkill
- subway-infoSkill
- systematic-debuggingSkill
- tui-designerSkill
- ui-designSkill
- video-editorSkill
- writing-plansSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 10 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈10
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
STOP: - Do not implement anything from the prompt
- Distinctive: line 37
Ask: "Would you like me to modify the prompt or create a variation?"
- Distinctive: line 46
Fix: ALWAYS ask 5-7 questions minimum, even for "obvious" requests
- Distinctive: line 76
Fix: Ask for format preference first.
- Very specific: line 82
Fix: Use smart adaptation in references/advanced-workflow.md → Skip truly redundant questions, emphasize domain/constraints
- Distinctive: line 83
Example: If user specifies "React 18 + TypeScript," skip tech stack question
- Distinctive: line 97
Ask: "Want a variation or different format?"
- Distinctive: line 108
Ask: "Want to modify the prompt or try a different format?"
- Distinctive: line 162
Read on-demand based on context:
- Distinctive: line 175
When generating prompts, incorporate these proven techniques as appropriate:
- Distinctive: line 196
Always include 2-3 high-quality examples for consistent behavior:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Prompt Factory - World-Class Prompt Powerhouse
Generate production-ready mega-prompts through intelligent questioning and comprehensive domain presets.
⚠️ CRITICAL CONSTRAINTS - READ FIRST
This skill generates PROMPTS only. It does NOT implement the work described in the prompt.
What This Skill DOES:
✅ Generate a comprehensive PROMPT (text document in chosen format) ✅ Ask 5-7 questions to understand requirements (MANDATORY - no skipping) ✅ Validate prompt quality before delivery ✅ Output a SINGLE prompt document with token count ✅ Provide the prompt ready to copy and use elsewhere
What This Skill DOES NOT Do:
❌ Implement the actual work (no code files, no diagrams, no APIs) ❌ Create architectural diagrams or technical implementations ❌ Write actual marketing campaigns or business strategies ❌ Build infrastructure or deploy anything ❌ Execute the prompt after generating it
Expected Workflow:
- User asks for help creating a prompt
- Skill MUST ask 5-7 questions (even if context seems obvious)
- User answers questions with specific details
- Skill generates ONE comprehensive prompt document
- Skill announces token count
- STOP - Do not implement anything from the prompt
- Ask: "Would you like me to modify the prompt or create a variation?"
Gotchas & Common Pitfalls
1. Skipping Questions
Danger: Agent rushes to prompt generation without asking questions
- Why it breaks: Vague prompts miss key context, leading to poor outputs when user applies them
- Fix: ALWAYS ask 5-7 questions minimum, even for "obvious" requests
- Example: User says "PM prompt for PRD" but doesn't specify domain, constraints, or team context
- Wrong: Jump to generate PM preset
- Right: Ask: domain, PRD type, team size, constraints, success criteria
2. Implementing Instead of Generating
Danger: Agent starts coding/building the thing the prompt describes
- Why it breaks: Scope creep; this skill only makes prompts, not implementations
- Fix: Generate the prompt → Stop → Offer variations only
- Example: Prompt describes "REST API for payments"
- Wrong: Start coding the API
- Right: Generate prompt → User takes prompt elsewhere to implement
3. Conflating with PROMPTS_FACTORY_PROMPT.md
Danger: User asks "make a prompt system for FinTech" and you generate one mega-prompt
- Why it breaks: That's a meta-prompt task (building a prompt builder), not a single role prompt
- Fix: Redirect to PROMPTS_FACTORY_PROMPT.md for domain-wide builders
- Difference:
- This skill: One prompt for one role (e.g., "FinTech PM") → ~5K tokens
- Meta-prompt: Entire FinTech system with 10-20 role presets → generate with PROMPTS_FACTORY_PROMPT.md
4. Token Count Inflation
Danger: Generated prompts exceed optimal ranges (core: 3-6K, advanced: 8-12K)
- Why it breaks: Oversized prompts are harder to use; undersized prompts lack specificity
- Fix: Apply optimization tips from
references/prompt-patterns.md - Monitor: Announce token count; flag if core >8K or advanced >15K
5. Generating Multiple Formats by Default
Danger: Generating all 4 formats when user asked for one
- Why it breaks: Token bloat; user only needs one format
- Fix: Ask for format preference first; only generate all formats if user requests
- Default: XML (optimal for LLM parsing)
6. Weak Question Flow
Danger: Asking 7 generic questions, not 5-7 contextually relevant ones
- Why it breaks: Questions don't validate assumptions; responses lack specificity
- Fix: Use smart adaptation in
references/advanced-workflow.md→ Skip truly redundant questions, emphasize domain/constraints - Example: If user specifies "React 18 + TypeScript," skip tech stack question
- Still ask: domain, task, constraints, success criteria (always validate)
Quick Start: Choose Your Path
Path 1: Quick-Start Preset (Fastest)
Use when: You need a prompt for a common role
- User says: "I need a prompt for [preset name]"
- Confirm the preset matches their need
- Customize any variables (optional)
- Generate → Deliver with token count
- Ask: "Want a variation or different format?"
Available Presets: 69 across 15 domains (see references/presets.md)
Path 2: Custom Prompt (5-7 Questions - MANDATORY)
Use when: Building a unique prompt from scratch
- Detect intent from user request
- MUST ask 5-7 questions with example answers (mandatory - even for "obvious" requests)
- Apply contextual best practices
- Validate quality → Deliver with token count
- Ask: "Want to modify the prompt or try a different format?"
Full 8-step workflow: See references/advanced-workflow.md
Expected Output Formats
Choose one or generate all:
- XML (default) — Optimal for LLM parsing with structured tags
- Claude — Claude system prompt format
- ChatGPT — Custom instructions format
- Gemini — Google Gemini format
- All — Generate all 4 formats (use sparingly - token heavy)
See references/examples.md for complete examples of each format.
Generation Modes
- Core (default) — Prompt + instructions + 2-3 examples (~4-6K tokens)
- Advanced — Core + testing scenarios + variations + optimization tips (~10K tokens)
Quality Validation (7-Point Gates)
Before delivery, validate:
- ✓ XML Structure valid (if XML format)
- ✓ Completeness — All questions incorporated
- ✓ Token Count optimal (core: 3-6K, advanced: 8-12K)
- ✓ No Placeholders — All
[...]filled - ✓ Actionable Workflow — Clear, executable steps
- ✓ Best Practices applied contextually
- ✓ Examples present (minimum 2)
If validation fails: Fix before delivery.
Token Count Announcement
After generating, announce:
- "Token Count: ~4,200 tokens (Core mode - within optimal range ✅)"
- "Token Count: ~10,500 tokens (Advanced mode - comprehensive ✅)"
- "Token Count: ~7,800 tokens (Warning: Higher than typical Core mode)"
Reference Files
Read on-demand based on context:
| File | When to Use | Contains |
references/presets.md | User asks for quick-start | 69 preset names, domains, use cases |
references/advanced-workflow.md | Custom prompt path (Path 2) | 8-step workflow, detailed questions, validation |
references/prompt-patterns.md | During generation | Best practices (OpenAI/Anthropic/Google), pattern library, template matching |
references/examples.md | Format clarification | 2-3 worked examples per format (XML, Claude, ChatGPT, Gemini) |
Modern Prompting Techniques (2025/2026)
When generating prompts, incorporate these proven techniques as appropriate:
Chain-of-Thought (CoT)
For reasoning tasks, instruct the model to think step-by-step:
<instructions>
Think through this step-by-step before giving your final answer.
Show your reasoning process explicitly.
</instructions>Structured Outputs
For prompts that need machine-parseable responses:
<output_format>
Respond ONLY with valid JSON matching this schema:
{"result": string, "confidence": number, "reasoning": string}
</output_format>Few-Shot Examples
Always include 2-3 high-quality examples for consistent behavior:
<examples>
<example>
<input>Schedule a meeting for next Tuesday</input>
<output>{"action": "create_event", "day": "next_tuesday", "title": "Meeting"}</output>
</example>
</examples>Role + Context + Task Structure (optimal for Claude)
<role>You are a senior {domain} specialist with 15 years of experience at top-tier firms.</role>
<context>You are helping {user_type} with {specific_situation}.</context>
<task>{specific_instructions}</task>
<constraints>{limitations_and_guardrails}</constraints>
<output_format>{format_specification}</output_format>Negative Instructions
Explicitly stating what NOT to do reduces unwanted behaviors:
<constraints>
- Do NOT include unsolicited advice beyond the specific question
- Do NOT use jargon without explanation
- Do NOT recommend third-party tools unless asked
</constraints>Model-Specific Formatting Notes
- Claude: Responds best to XML tags; prefers explicit thinking steps; use
<thinking>for CoT - GPT-4.1: Responds well to markdown; system prompt is highly influential
- Gemini 2.5: Responds well to numbered lists; excels with explicit JSON schema specs
Common Questions
"Do you implement the prompt after generating it?" → No. This skill makes prompts only. Once generated, you take it elsewhere to implement or use with another LLM.
"Can I get a variation of this prompt?" → Yes. Ask for "concise version," "more detailed," "different tone," or "different format." We'll modify the prompt (not reimplement it).
"What if the generated prompt doesn't work?" → We can refine it. Share what didn't work, and we'll adjust specific sections, not rebuild from scratch.
"Why so many questions?" → Good prompts need context. 5-7 questions ensure we capture domain specifics, constraints, and success criteria—making the final prompt much more effective.
Ready to generate a world-class prompt? Let's start with a preset name or custom request.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

