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 you activate inside any AI chat session. It instructs the AI to act as an adversarial reviewer — finding gaps, weak logic, missing constraints, and shaky fixes in whatever you submit. No praise, no style polish: just concrete problems with structural fixes.

Quickstart

Get your first adversarial review in under two minutes.

When to Use It

See which artifacts benefit most from adversarial review.

Two-AI Ping-Pong

Run alternating review passes across two independent AI chats.

Protocol Reference

Explore all input blocks, output fields, and review rules.

How it works

PA-PVP Mini gives any AI chat a strict behavioral contract. You paste the protocol text into a session, then submit your artifact. The AI returns a structured list of findings — each one anchored to your material, assigned a severity, and paired with a concrete fix.
1

Download the protocol

Grab pa_pvp mini v2.3.txt from the GitHub repository. This is the single file that powers everything.
2

Activate in your AI chat

Paste the protocol text into a new chat session (or upload the file if the platform supports attachments). The AI is now operating under the adversarial review contract.
3

Submit your artifact

Send the content you want reviewed. Label it clearly for best results:
ORIGINAL_ARTIFACT:
[your material here]
4

Receive structured findings

The AI returns Round 1 — a list of findings with severity levels, evidence anchored to your artifact, and a concrete fix for each problem.
5

Continue with a second AI (optional)

Copy the output and paste it into a second AI chat running the same protocol. The second AI reviews both your artifact and the first review independently — the ping-pong begins.

What makes it different

No praise, no fluff

The protocol explicitly prohibits narrative commentary and encouragement. Every output item must be observable, specific, and actionable.

Every finding has a fix

A finding with no valid structural fix is excluded. You never receive a problem without a path to resolving it.

Provenance tracking

Finding IDs, fix IDs, and derived_from links trace each issue across multiple review rounds.
PA-PVP Mini does not prove that something is correct. It is a review aid, not an authority. The quality of findings depends on the AI model reading the protocol. Read the Limits and Caveats page before relying on it for high-stakes decisions.

Build docs developers (and LLMs) love