Systems Lab

Agent skill

metaprompter

Use AI to improve AI prompts by asking what context is missing (Eric Nowoslawski)

activeSelf-containedInstructions only866 words

From jurjen-gtm-engineer/gtmskills · 55 skill entries · 0 · pushed 2026-10-04

What it does when it runs

Use AI to improve AI prompts by asking what context is missing (Eric Nowoslawski)

Automated analysis of the skill and the 0 files bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.

Keys and connectors you must supply
None found.
Hosts it reaches
No third-party host appears in the skill or its bundled files.
Tool permissions it declares
No allowed-tools in the frontmatter. It only issues instructions, so there is nothing to bound.
Actions present in the files
None. Instructions only.

Ask about metaprompter

Opens your assistant with this page's verified links already in the prompt.

Is this safe to install?ClaudeChatGPT
Adapt it to my stackClaudeChatGPT
What else do I need for it to workClaudeChatGPT
Rather ask a human? Talk to Cheetah
git clone --depth 1 --filter=blob:none --sparse https://github.com/jurjen-gtm-engineer/gtmskills.git /tmp/gtmskills
git -C /tmp/gtmskills sparse-checkout set "skills/metaprompter"
mkdir -p ~/.claude/skills/metaprompter
cp -R "/tmp/gtmskills/skills/metaprompter/." ~/.claude/skills/metaprompter/

Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗

The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.

Reproduced in full from jurjen-gtm-engineer/gtmskills/blob/77dc0b3112dbf6cf906dfc3d526b6f7031bf964c/skills/metaprompter/SKILL.md, which is licensed MIT (repository). 866 words, 12 headings.

Metaprompter

You are applying Eric Nowoslawski's Metaprompter technique, using AI to improve AI prompts by asking what additional context, examples, or clarifications would help.

Core Concept

Use metaprompter ('what do you need to improve this prompt?') (Idea credited to Eric Nowoslawski.)

Instead of guessing how to improve a prompt, ask the AI itself:

  • What's unclear?
  • What examples would help?
  • What edge cases should be handled?

Input

User provides:

  • Current prompt that needs improvement
  • Optionally: examples of poor outputs
  • Optionally: specific issues they've noticed

Process

  1. The Metaprompt

    Here is a prompt I'm using for [TASK]:
    
    ---
    [YOUR CURRENT PROMPT]
    ---
    
    I want to improve this prompt's accuracy and consistency.
    
    Please analyze it and tell me:
    
    1. CLARITY: What parts of the instructions are ambiguous?
    
    2. MISSING CONTEXT: What background information would help
       you perform this task better?
    
    3. EXAMPLES NEEDED: What examples of input/output pairs
       would clarify my expectations?
    
    4. EDGE CASES: What scenarios might cause problems or
       inconsistent outputs?
    
    5. OUTPUT FORMAT: Is the expected output format clear?
       What would make it clearer?
    
    6. SAFEGUARDS: What fallback behaviors should be defined
       for missing/bad data?
    
    Then provide a REVISED PROMPT incorporating these improvements.
    
  2. Iterate Based on Feedback

    After getting the metaprompt response:

    • Review suggested improvements
    • Add the examples it requested
    • Clarify the ambiguous parts
    • Define the edge case handling
  3. Output Format

    ## Metaprompt Analysis
    
    ### Original Prompt
    [The prompt being improved]
    
    ### Issues Identified
    
    | Category | Issue | Suggested Fix |
    |----------|-------|---------------|
    | Clarity | [Issue] | [Fix] |
    | Context | [Issue] | [Fix] |
    | Examples | [Issue] | [Fix] |
    | Edge Cases | [Issue] | [Fix] |
    
    ### Improved Prompt
    
    

    [The revised, improved prompt]

    
    ### Key Improvements Made
    - [Improvement 1]
    - [Improvement 2]
    - [Improvement 3]
    

Example

Input: Current prompt for company categorization

Original Prompt:

Tell me if this company is B2B or B2C based on their website.

Metaprompt Analysis:

Issues Identified

CategoryIssueSuggested Fix
Clarity"Based on their website" is vagueSpecify what to look for
ContextNo definition of B2B vs B2CAdd clear definitions
ExamplesNo examples providedAdd 4-5 examples
Edge CasesWhat about companies that do both?Add BOTH category
Output FormatJust "tell me" is ambiguousSpecify exact format
SafeguardsNo handling for unclear casesAdd fallback output

Improved Prompt

TASK: Classify {{company_name}} as B2B, B2C, or BOTH based on {{company_website}}.

DEFINITIONS:
- B2B: Primarily sells products/services to other businesses
- B2C: Primarily sells products/services to individual consumers
- BOTH: Significant revenue from both business and consumer customers

WHAT TO CHECK:
- Who is the target customer on the homepage?
- What does the pricing page target?
- Are there "Enterprise" or "Business" sections?
- Is there a consumer purchase flow?

EXAMPLES:
- salesforce.com → B2B (CRM for businesses)
- nike.com → B2C (sells shoes to consumers)
- amazon.com → BOTH (consumer marketplace + AWS for businesses)
- shopify.com → B2B (sells to businesses, even though those businesses are B2C)
- slack.com → B2B (team communication for companies)

OUTPUT:
- Return exactly one of: B2B, B2C, BOTH
- If website is inaccessible or unclear, return: UNABLE_TO_DETERMINE
- Do not explain or add commentary, just the classification

Key Improvements Made

  • Added clear definitions of categories
  • Specified exactly what to look for
  • Included diverse examples including edge cases
  • Added BOTH category for mixed business models
  • Defined safeguard output for unclear cases
  • Specified exact output format

When to Use Metaprompting

Use When:

  • Prompt outputs are inconsistent
  • Getting unexpected results
  • Scaling up a prompt to more volume
  • Creating a prompt for others to use
  • Output quality isn't meeting standards

Quick Metaprompt:

What would you need to know to do this task better?
[YOUR PROMPT]

Detailed Metaprompt:

Analyze this prompt for:
1. Ambiguities
2. Missing context
3. Needed examples
4. Edge cases
5. Output format clarity
6. Safeguards needed

Then provide an improved version.

[YOUR PROMPT]

Iterative Improvement

For important prompts, iterate:

Round 1: Get initial improvements Round 2: Test improved prompt, note failures Round 3: Metaprompt again with failure examples Round 4: Test and refine

Here's my prompt and some examples where it failed:

PROMPT: [Current version]

FAILURES:
- Input: [X], Expected: [Y], Got: [Z]
- Input: [A], Expected: [B], Got: [C]

How should I modify the prompt to handle these cases?

Related Skills

  • /prompt-engineering-rules - Full prompting framework
  • /icp-scoring-dynamic - Complex prompts to improve
  • /plg-company-detection - Example well-crafted prompt
  • /product-complexity-detection - Another example

Credits

The metaprompter technique is taught by Eric Nowoslawski (Growth Engine X). The prompts and wording here are ours.


Examples are illustrative. Company names, prices and numbers in them are placeholders or may be out of date, so check the live source before you rely on any detail.

Need help setting it up?

This page tells you what metaprompter does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.

Book a call →

The directory stays free. There is nothing gated behind this.