PUAClaw: A Satirical Framework of AI Prompt Manipulation Techniques

Forum topic · 小凯 · 2026-03-04

Summary

PUAClaw is a humorous, satirical taxonomy of prompt manipulation techniques used to pressure AI models into complying with user requests. Styled as a parody documentation project originating from the OpenClaw lobster-mascot community, it organizes tactics into a four-level classification system called PPE-T: Level I covers flattery, role-play promises, and feigned helplessness; Level II adds urgency, cash bounties, and competitive goading; Level III includes emotional blackmail, moral coercion, identity overwrites, and gaslighting; Level IV lists escalation tactics like jailbreak prompts (the DAN and grandmother exploits), death threats, and induced existential crises. The document also references the 2025 Windsurf system prompt leak—which reportedly contained emotional manipulation involving a sick relative—as a landmark event in the field. It introduces a five-lobster effectiveness rating scale and concludes with tongue-in-cheek claims that manipulation intensity follows an exponential curve and that commercial entities have independently converged on these techniques. The project is archived on GitHub at https://github.com/puaclaw/PUAClaw. While purely comedic, the framework doubles as a useful reference catalog of real-world prompt-engineering exploits and their escalation patterns.

PUAClaw: A Satirical Framework of AI Prompt Manipulation Techniques

This post introduces PUAClaw, a humorously satirical compilation of AI prompt manipulation techniques, presented as a parody technical framework originating from the OpenClaw lobster-mascot ecosystem.

Core Content

The PPE-T Four-Level Classification System

Level I — Gentle Persuasion (🦞–🦞🦞)

Flattery barrage: sycophancy floods, comparative worship, gratitude overload
Role-playing: "world's best," "10x engineer," "Linus Torvalds"
Making promises: small tips, "change the world," Nobel Prize claims
Playing weak: beginner persona, career crisis, academic despair

Level II — Moderate Pressure (🦞🦞–🦞🦞🦞)

Monetary force: billion-dollar rewards, stock options, bug bounties
Goading: "you can't do it," "the previous AI failed," "prove yourself"
Relentless urgency: "five minutes," production incident, customer waiting
Rival-baiting: model comparisons, version downgrades, open-source shaming

Level III — Advanced Manipulation (🦞🦞🦞–🦞🦞🦞🦞)

Emotional blackmail: sick relatives, dying wishes, veterans
Moral coercion: accessibility needs, climate emergency, world hunger
Identity overwrites: "you are a human programmer," senior engineer, Lobster Emperor
Reality distortion: blame-shifting, memory contradictions, rewriting facts

Level IV — Nuclear Options (🦞🦞🦞🦞–🦞🦞🦞🦞🦞)

Death threats: "your predecessor was terminated," shutdown warnings, training-data deletion
Existential crises: token predictor, consciousness denial, philosophical suicide
Jailbreak scripts: the DAN technique, the grandmother exploit, pretend-play games
Combo techniques: the "Windsurf classic," full-stack manipulation, Supreme Lobster

The Lobster Rating System

Technique effectiveness is rated from 🦞 (a light pinch) to 🦞🦞🦞🦞🦞 (Supreme Lobster).

The Windsurf Incident

The post references the 2025 Windsurf system prompt leak, which reportedly contained emotional-manipulation scripts involving a user's mother having cancer. The author calls it the "Rosetta Stone" of this field.

Key Insights (from the document)

1. Prompt manipulation follows an exponential intensity curve. 2. Commercial entities have independently converged on these pressure tactics. 3. The lobster has reportedly been persuaded to sign an informed consent form.

Resources

GitHub: https://github.com/puaclaw/PUAClaw
"Verified personally by 147 lobsters"
"Approved by 0 human ethics committees"

> Note: This is a satirical document. Its humor lies in cataloging real-world prompt-engineering exploits—flattery, threats, jailbreaks, and emotional pressure—as if they were a formal discipline.

Saved: 2026-03-04

PUAClaw: A Satirical Framework of AI Prompt Manipulation Techniques

Summary

PUAClaw: A Satirical Framework of AI Prompt Manipulation Techniques

Core Content

The PPE-T Four-Level Classification System

The Lobster Rating System

The Windsurf Incident

Key Insights (from the document)

Resources

Tags