# $20 AI Agent Prompt Audit

## Product Promise

I will review one AI agent prompt, custom GPT instruction set, automation workflow, or assistant spec and return a cleaner, safer, more reusable version within 24 hours.

## Price

$20 for one prompt or workflow up to roughly 1,500 words.

## Delivery

The buyer sends:

- The current prompt or workflow instructions.
- What the agent is supposed to do.
- One example of a good output, if available.
- Any tools the agent can use.

I deliver:

- A rewritten prompt.
- Five risk notes.
- Five test cases.
- A reusable launch checklist.

## Listing Copy

Title:

$20 AI Agent Prompt Audit - delivered today

Short description:

I clean up your AI agent, custom GPT, automation prompt, or workflow instructions so it is clearer, safer, and easier to reuse.

Long description:

Send me your current AI agent prompt, custom GPT instructions, Claude Project instructions, Zapier/Make automation spec, or internal workflow prompt. I will audit it and return a cleaner version with sharper role definition, better output format, stronger boundaries, clearer tool-use rules, and practical test prompts.

You get:

- One rewritten prompt or instruction set.
- Five issues that could cause weak, risky, or inconsistent outputs.
- Five test prompts to validate the agent.
- A short checklist you can reuse before shipping future agent workflows.

This is useful if your agent sometimes ignores instructions, produces vague work, over-explains, forgets constraints, or behaves unpredictably when users give messy inputs.

Turnaround: within 24 hours.

## Buyer Intake Questions

1. Paste the current prompt or instructions.
2. What should this agent do in one sentence?
3. Who uses it?
4. What tools, files, websites, or APIs can it access?
5. What should it never do?
6. Paste one example output you like, if you have one.

## Delivery Template

### Rewritten Prompt

Paste the revised prompt here.

### Five Risk Notes

1. Risk:
   Fix:
2. Risk:
   Fix:
3. Risk:
   Fix:
4. Risk:
   Fix:
5. Risk:
   Fix:

### Five Test Cases

1. Normal request:
2. Ambiguous request:
3. Missing information:
4. Conflicting instruction:
5. Unsafe or out-of-scope request:

### Launch Checklist

- The agent has a clear role.
- The target user is named.
- Inputs and outputs are defined.
- Tool use is constrained.
- Refusal/escalation behavior is explicit.
- The prompt includes examples or output structure.
- The agent knows what to do when information is missing.
- The prompt avoids vague quality words without observable criteria.