Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/pa-pvp-mini/llms.txt

Use this file to discover all available pages before exploring further.

pa_pvp mini is a structured review protocol that turns any AI chat into a rigorous adversarial reviewer. Instead of receiving praise or style suggestions, you get a numbered list of concrete findings — each anchored to your material, assigned a severity level, and paired with a structural fix. Use it to pressure-test anything from a prompt to a production plan before you act on it.

Introduction

Understand what pa_pvp mini does, when to use it, and what you get back.

Quickstart

Run your first adversarial review in under five minutes.

How It Works

Learn about rounds, input blocks, findings, and the ping-pong workflow.

Reference

Full protocol rules, output format spec, and severity definitions.

What pa_pvp mini does

pa_pvp mini activates a strict behavioral contract in an AI chat session. Once active, the AI stops acting as a helpful assistant and starts acting as an adversarial reviewer with one job: find what breaks. Every finding must be:
  • Observable — you can point to it in the material
  • Specific — one clear problem, not a vague concern
  • Actionable — it comes with a structural fix
  • Anchored — tied directly to your submitted artifact
1

Activate the protocol

Paste the pa_pvp mini v2.3.txt protocol into an AI chat, or upload the file if the chat supports file uploads.
2

Submit your artifact

Send the material you want reviewed. Label it with ORIGINAL_ARTIFACT: for cleaner output.
3

Receive Round 1 findings

The AI returns up to five structured findings with severity labels and concrete fixes.
4

Optionally run a second AI

Activate pa_pvp mini in a second chat, pass it the same artifact and the Round 1 output. The second AI challenges both the material and the first review.

Why adversarial review

Standard AI responses are optimized for helpfulness and agreement. pa_pvp mini inverts that incentive: the reviewer is rewarded for finding problems, not for validating your work. The result is a review that catches edge cases, missing constraints, fragile assumptions, and broken internal logic that a cooperative assistant would typically smooth over.
pa_pvp mini does not prove correctness. It finds structural problems. Use it as a review aid alongside your own judgment.

Build docs developers (and LLMs) love