| name | prompt-library |
| description | Organizes and documents reusable prompt libraries with categories, variables, and quality scoring. Use when building a structured collection of AI prompts for your business. |
| allowed-tools | Read Write Glob |
| metadata | {"author":"matthewhitcham","version":"1.0"} |
Prompt Library
When to Use This Skill
Use this skill when you need to:
- Build an organized library of reusable AI prompts for business tasks
- Document prompts with variables, examples, and usage instructions
- Create a quality scoring system for prompt effectiveness
- Standardize prompt usage across a team or workflow
DO NOT use this skill for writing individual prompts, prompt engineering tutorials, or AI tool comparisons. This is for organizing and documenting a collection of prompts.
Core Principle
A PROMPT LIBRARY IS A BUSINESS ASSET — EVERY PROMPT SHOULD BE DOCUMENTED, TESTED, AND VERSIONED SO ANYONE CAN USE IT AND GET CONSISTENT RESULTS.
Phase 1: Brief
Required Inputs
| Input | What to Ask | Default |
|---|
| Business type | "What does your business do?" | No default — must be provided |
| AI tools used | "Which AI tools do you use? Claude, ChatGPT, Gemini, other?" | Claude |
| Key use cases | "What do you use AI for most? Content, email, research, analysis?" | No default — list at least 5 |
| Team usage | "Will this library be used by just you or shared with a team?" | Solo use |
| Current prompts | "Do you have any prompts you already reuse regularly?" | A few informal ones |
| Storage preference | "Where should the library live? Notion, Google Docs, markdown files?" | Markdown files |
GATE: Confirm the brief before designing the library structure.
Phase 2: Design the Library
Category Structure
Organize prompts by business function:
prompt-library/
├── content/
│ ├── blog-post-draft.md
│ ├── social-media-caption.md
│ └── email-newsletter.md
├── marketing/
│ ├── ad-copy.md
│ ├── landing-page-copy.md
│ └── lead-magnet-outline.md
├── operations/
│ ├── meeting-summary.md
│ ├── process-documentation.md
│ └── project-brief.md
├── finance/
│ ├── invoice-reminder.md
│ └── expense-report.md
├── customer/
│ ├── support-response.md
│ ├── review-reply.md
│ └── onboarding-email.md
└── research/
├── competitor-analysis.md
└── market-research.md
Prompt Card Template
Every prompt in the library uses this format:
# [Prompt Name]
**Category:** [Business function]
**AI Tool:** [Claude / ChatGPT / Any]
**Version:** [1.0]
**Last tested:** [Date]
**Quality score:** [1-5]
## Purpose
[One sentence: what this prompt produces]
## Variables
- {{VARIABLE_1}}: [Description, example]
- {{VARIABLE_2}}: [Description, example]
## Prompt
[The full prompt text with {{VARIABLES}} marked]
## Example Output
[A sample of what good output looks like]
## Tips
- [Usage tip 1]
- [Usage tip 2]
## Changelog
- v1.0 — Initial version
GATE: Confirm the structure and template before populating the library.
Phase 3: Populate
Prompt Writing Rules
- One prompt, one purpose — do not combine tasks. A prompt that writes AND edits AND formats is three prompts.
- Variables are explicit — use
{{DOUBLE_BRACES}} for every input the user must provide. Never leave implicit assumptions.
- Include constraints — word count, tone, format, what to avoid. Constraints improve output quality.
- Show the ideal output — include one example of what a good response looks like. This sets the quality bar.
- Version everything — when you improve a prompt, increment the version and note what changed.
Quality Scoring Rubric
| Score | Label | Criteria |
|---|
| 5 | Production-ready | Consistent high-quality output with minimal editing needed |
| 4 | Reliable | Good output 80%+ of the time, occasional light editing |
| 3 | Usable | Produces a solid starting point but requires human refinement |
| 2 | Needs work | Inconsistent results, prompt needs improvement |
| 1 | Experimental | Untested or unreliable, keep for iteration |
Prompt Iteration Process
- Write the initial prompt (v1.0)
- Test with 3 different inputs
- Score the outputs
- Identify weaknesses (too long, wrong tone, missing elements)
- Revise the prompt (v1.1)
- Retest and rescore
- Repeat until score reaches 4+
Phase 4: Polish
1. Library Maintenance
- Monthly review: Test top 10 most-used prompts with current AI model versions
- Quarterly audit: Archive prompts not used in 90 days, add new ones for emerging needs
- Version control: Never overwrite — always create a new version and note changes
- Usage tracking: Note which prompts get used most and which produce the best results
2. Team Sharing Guide (if applicable)
- Store in a shared location (Notion, shared folder, repo)
- Create a quick-start guide: how to find prompts, how to use variables, how to submit new prompts
- Designate a prompt librarian (or rotate the role)
- Set up a feedback mechanism: "This prompt worked / didn't work because..."
3. Quality Checklist
## Prompt Library Checklist
- [ ] Categories cover all major business functions
- [ ] Every prompt uses the standard card template
- [ ] All variables are marked with {{DOUBLE_BRACES}} and described
- [ ] Each prompt includes at least one example output
- [ ] Quality scores are assigned and up to date
- [ ] Prompts are versioned with changelogs
- [ ] Library is organized in a searchable structure
- [ ] Monthly review cadence is scheduled
- [ ] Top prompts are tested with current AI model versions
- [ ] Archive process exists for unused prompts
Example
Prompt card:
# Social Media Caption
**Category:** Content
**AI Tool:** Claude
**Version:** 1.2
**Last tested:** 2026-02-15
**Quality score:** 4
## Purpose
Writes a social media caption for Instagram or LinkedIn from a topic and key message.
## Variables
- {{PLATFORM}}: Instagram or LinkedIn
- {{TOPIC}}: The subject of the post
- {{KEY_MESSAGE}}: The one point you want to make
- {{TONE}}: casual, professional, motivational, witty
- {{CTA}}: What you want the reader to do
## Prompt
Write a {{PLATFORM}} caption about {{TOPIC}}.
Key message: {{KEY_MESSAGE}}
Tone: {{TONE}}
Include a call to action: {{CTA}}
Requirements:
- Under 150 words for Instagram, under 200 for LinkedIn
- Open with a hook (question, bold statement, or surprising fact)
- Use line breaks for readability
- End with the CTA
- No hashtags in the body (I will add them separately)
## Example Output
"Most solopreneurs spend 6 hours a week on invoicing.
Six. Hours.
That is an entire workday every month spent on paperwork instead of revenue.
I cut that to 45 minutes by automating three things:
→ Invoice generation from project completion
→ Automatic payment reminders
→ Expense categorization
The tools cost $30/month total. The time savings are worth $1,200.
Drop a 🔥 if you want the exact setup."
## Changelog
- v1.2 — Added "no hashtags" constraint, improved hook instruction
- v1.1 — Added platform-specific word count limits
- v1.0 — Initial version
Anti-Patterns
- Hoarding untested prompts — a library of 50 untested prompts is a junk drawer. Test and score before adding.
- No variables — hardcoded details make prompts single-use. Extract every changeable element into a variable.
- No examples — without a sample output, quality is subjective. Always show what good looks like.
- Never updating — AI models change. A prompt that worked 6 months ago may not work as well today. Review regularly.
- Overcomplicating — a prompt that requires 15 variables is too complex. Break it into multiple simpler prompts.
Recovery
- Too many prompts to organize: Start by categorizing the 10 you use most. Archive everything else and add as needed.
- Prompts produce inconsistent results: Add more constraints (word count, format, tone, examples). Constraints reduce variance.
- Team members not using the library: The library is too hard to access or search. Simplify structure and create a quick-start guide.
- Quality scores are all low: The prompts may be too broad. Narrow the purpose of each prompt to one specific output.
