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 plain-text adversarial review protocol you paste into any AI chat. It is not a code library, not an app, and requires no installation. When activated, it sets a strict behavioral contract: the AI becomes an adversarial reviewer whose only job is to find what breaks, what is weak, what is missing, and what structural fix should exist instead. It produces a short, actionable list of findings — nothing more.

What it reviews

pa_pvp mini works on any artifact that carries internal structure: steps, claims, constraints, rules, dependencies, expected behavior, or logic that can be challenged. The following types of material are all valid inputs:
  • Ideas that sound plausible but may have gaps
  • Plans and proposals
  • Decision memos
  • Processes and checklists
  • Policies and rule sets
  • Prompts and AI workflows
  • Structured arguments
  • Research outlines
  • Technical specs
  • Lists of bugs, fixes, or refactors
  • AI-generated answers you want to pressure-test
The protocol is intentionally broad. If the material has structure that can be broken, pa_pvp mini can review it.

When to use it

pa_pvp mini is most effective when the material has internal logic that can be checked against itself — steps that must be in order, constraints that must hold, claims that must be supported by evidence within the artifact.
Use pa_pvp mini when:
  • Something looks plausible but you do not want to trust it yet
  • An AI gave you an answer and you want to pressure-test it before acting on it
  • You have proposed fixes and want to know which ones actually hold up
  • You wrote a plan, prompt, procedure, proposal, or spec and want a stricter review than encouragement
  • You want a second AI to challenge the first AI’s review rather than simply agree with it
  • You have an existing structured review from any source and want it audited against the original artifact

What you get back

Every review returns a short list of findings. Each finding is built from four required parts:
  • What breaks — a single concise statement of the structural problem
  • Evidence — an observable reason anchored directly to the artifact under review
  • Severity — one of HIGH, MED, or LOW
  • Fix — a concrete structural change that addresses the problem
Findings with no valid fix are excluded entirely. A fix is only valid if it changes structure, behavior, constraints, decision rules, flow, validation, or operational clarity. Pure wording changes do not qualify unless they remove ambiguity that alters how the artifact works.
The output never includes praise, encouragement, or narrative commentary. The goal is a list of review items that can be checked, challenged, merged, or applied — not a set of opinions.

Key constraints

Apply pa_pvp mini findings with judgment. Prefer findings that have clear evidence or that are confirmed across multiple review rounds before making structural changes.
pa_pvp mini has hard limits you should keep in mind:
  • Does not prove correctness. A clean review means no structural findings were strong enough to include. It does not mean the artifact is correct or complete.
  • Does not find every problem. The protocol caps findings at five per round, keeping only the strongest. Real issues beyond that count may not surface in a single pass.
  • Model-dependent. Output quality varies with the AI model reading the protocol. The same protocol text may produce different findings in different models or sessions.
  • Review aid, not authority. pa_pvp mini surfaces candidates for structural improvement. Deciding whether to act on a finding is always a human judgment.

Quickstart

Activate the protocol, submit your first artifact, and read a real finding in minutes.

How it works

Understand rounds, provenance, severity levels, and the two-AI ping-pong mode.

Build docs developers (and LLMs) love