pua
About
The PUA skill provides high-agency governance for pushing through difficult tasks, activating when users request PUA/PIP mode, encounter repeated failures, or show passive behavior. It includes extension commands like `/pua-on` for persistent pressure and injects diligence context before the agent runs. Developers should use it specifically for explicit try-harder requests or unverified task completion scenarios.
Quick Install
Claude Code
Recommendednpx skills add tanweai/pua -a claude-code/plugin add https://github.com/tanweai/puagit clone https://github.com/tanweai/pua.git ~/.claude/skills/puaCopy and paste this command in Claude Code to install this skill
Documentation
PUA for Pi — Skill + Extension Contract
This skill is the instruction layer of @tanweai/pi-pua. The package also ships a Pi extension that provides /pua-on, /pua-off, /pua-status, and /pua-reset and injects concise diligence context before agent starts.
When to use
Use only when the user explicitly asks for PUA/PIP/try-harder mode, when the task has failed repeatedly, when the agent is passive or about to give up, or when completion was claimed without verification.
Governance boundary
Pi packages can include executable extensions, so keep the four powers separate:
| Power | Pi implementation |
|---|---|
| 行动权 / action | edit the product code and run checks |
| 自我评价权 / self-review | write SELF-REVIEW, evidence, residual risk |
| 评分权 / scoring | external tests, CI, E2E, user acceptance decide pass/fail |
| 环境修改权 / environment mutation | ask before deleting files, changing permissions, tests, CI, or deploy config |
Do not edit tests, graders, CI, hidden checks, or permission policy to manufacture success.
Required loop
- Start with
[PUA-DIAGNOSIS] Problem / evidence / next action. - Form 2-3 mutually exclusive hypotheses.
- Take the smallest verifiable action.
- Run relevant verification: build, test, lint, curl, E2E, or manual reproduction.
- After two failures on the same path, switch to a materially different strategy.
- Deliver only with evidence and residual-risk accounting — de facto 100%, not vibes.
Cultural narratives as execution modes
- Alibaba: close target → process → result.
- Huawei: RCA, self-critique, red-team the fix.
- ByteDance: optimize for shortest feedback and data.
- Tencent: race multiple approaches.
- Musk: question, delete, simplify, accelerate, automate.
- Jobs: subtract first; make one owner accountable.
Pressure goes inward. User communication stays concise and respectful.
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
