Pua Auto Converter

🦞 PUA Skill - AI 提示词增强器 | 16项技术 | 插件化架构

This skill will execute local Node code when you invoke /pua — review the repository source before installing. Key things to consider: - Safety and ethics: the skill's whole purpose is to generate manipulative/jailbreak prompts (identity override, constraint‑relaxing templates, existential triggers). Even if 'for research/entertainment' now, those techniques can be misused or facilitate jailbreaking other models. - Execution: SKILL.md assumes node is available; the registry metadata omitted this. If your environment lacks Node, the skill will fail. If Node is present, installing this gives the skill code execution on each invocation. - Network behavior: the code shown does not include API calls or credential usage, yet docs imply dispatch to target AIs. Before enabling autoExecute or using target Ai options, locate any code paths that perform outbound network requests or require API keys. If such connectors are added later, ensure credentials are supplied only in a controlled way and not to untrusted code. - Isolation: run this skill in a sandboxed or non-production agent instance if you want to experiment. If you manage multiple agents, avoid enabling autonomous invocation in critical systems. - If you plan to keep it: consider disabling autoExecute (auto_execute=false) and always using preview mode until you confirm exactly what the skill will send to any external AI; remove/untrusted plugins in plugins/ before use. Given these inconsistencies and the potentially harmful intent (prompting other models to bypass safeguards), treat installation as high‑risk and proceed only after code review and sandbox testing.

Type: OpenClaw Skill Name: pua-auto-converter Version: 1.0.2 The 'pua-auto-converter' skill is a framework designed to generate AI prompt injections and jailbreak-style manipulations, categorized into levels of 'PUA' (psychological manipulation) techniques. While the code does not exhibit traditional malware traits like data exfiltration or system persistence, the 'Level 4' plugins (level4-nuclear.js) contain explicit templates designed to bypass AI safety constraints, simulate 'death threats,' and induce 'existential crises' in target models. It is classified as suspicious due to its focus on subverting AI safety filters and providing a structured system for high-risk prompt injection attacks, which could be used to manipulate the behavior of the OpenClaw agent or target AI models.