Single-AI mode is the simplest way to use PA-PVP Mini. You activate the protocol once inside a chat session, submit your artifact, and receive a structured list of findings. You can run multiple review rounds inside the same chat by re-submitting your original artifact alongside the previous round’s output — no second AI needed.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.
Session flow
Paste the protocol into a fresh chat
Open a new chat session with your AI of choice. Paste the full contents of
pa_pvp mini v2.3.txt into the chat to activate PA-PVP Mini as the behavioral contract for that session. The AI will not return praise, narrative commentary, or style suggestions — only structured findings with evidence and fixes.Submit your artifact with the ORIGINAL_ARTIFACT label
Send the material you want reviewed. Label it explicitly so the protocol treats it as the primary source of truth:The artifact can be an idea, plan, prompt, procedure, code fix, proposal, spec, or AI-generated output — anything with internal structure the protocol can pressure-test.
Receive Round 1 FINDINGS
The AI returns a Round 1 output. Every finding is labeled
INDEPENDENT and includes:- what_breaks — one concise statement of the problem
- evidence — observable reason anchored to your artifact
- fix — a structural change that addresses the problem
- severity —
HIGH,MED, orLOW
END OF PROTOCOL.Continue to Round 2 by re-submitting with PREVIOUS_ROUND_OUTPUT
To run a second round, send a new message that contains both your original artifact and the full Round 1 output:The AI runs a fresh independent review of the artifact first, then cross-checks Round 1’s findings, and returns a Round 2 output that includes a
MERGE STATUS block summarising what was confirmed, challenged, or newly found.How round detection works
PA-PVP Mini detects the current round automatically from the blocks you supply. IfPREVIOUS_ROUND_OUTPUT is absent, the AI uses Round 1 behavior. If PREVIOUS_ROUND_OUTPUT is present, it switches to Round 2+ behavior — running an independent review first, then auditing the prior output. You never need to number rounds manually unless you want to override detection with an explicit CURRENT_ROUND: block.
Optional: MEMORY_NOTE
MEMORY_NOTE is an optional short carry-over context block you can include alongside PREVIOUS_ROUND_OUTPUT. Use it to pass a brief note between sessions — for example, a constraint you added after Round 1, or context that isn’t captured in the artifact itself. Like PREVIOUS_ROUND_OUTPUT, it is treated as audit material, not as source authority.
If you want a second, independent AI to challenge the same artifact and audit the first AI’s findings, see Two-AI Mode. Each AI operates in its own chat with no shared memory, making disagreements between them more meaningful.