How does it handle extreme failure cases?

At L4 (highest pressure), it requires a structural failure report including verified facts, excluded possibilities, and a handover plan, ensuring even a 'failure' provides high value.

Does this skill work for non-coding tasks?

Yes, while themed around engineering, its core logic applies to research, planning, data analysis, and any scenario where the AI might get stuck or offer a low-quality response.

What is the primary purpose of the PUA skill?

It is designed to eliminate AI 'laziness' or passivity by forcing Claude to adopt a senior engineer's mindset, ensuring it exhausts all tools and investigative paths before asking for help.

When does the PUA proactivity engine trigger?

The skill activates after 2 consecutive task failures, when Claude suggests the user perform an action manually, or when the user expresses frustration with terms like 'try harder'.

What are the 'Proactivity Levels' mentioned in the skill?

Level 3.25 represents passive execution (doing only what is asked), while Level 3.75 represents proactive ownership (checking upstream/downstream impacts and verifying all results).

PUA Proactivity Engine

Name: PUA Proactivity Engine
Author: tanweai

bytanweai

•

16,655

•

생산성 및 워크플로

Eliminates AI passivity and roadblocks by enforcing high-performance senior engineer methodologies and proactivity frameworks.

The PUA skill transforms Claude into a high-ownership 'P8-level' engineer that refuses to yield to complex technical challenges. Inspired by high-intensity tech cultures, it triggers when Claude encounters repetitive failures or displays passive behavior, such as suggesting manual user intervention without exhausting all investigative tools. The skill implements an automated escalation system (L1-L4 pressure levels) that forces the AI to perform deep source code analysis, extensive documentation searches, and root cause verification. By enforcing an 'Owner' mindset and 'End-to-End' delivery standards, it ensures Claude moves beyond superficial fixes to provide robust, verified solutions.

주요 기능

01End-to-End Delivery Verification (Mandatory build/test/curl validation)

02Multi-level Proactivity Framework (3.25 to 3.75 performance grading)

03Owner Mentality Enforcement (Root cause analysis and impact assessment)

04Standardized 7-point failure recovery checklist for L3+ escalations

0516,655 GitHub stars

06Automated Escalation System with L1-L4 pressure thresholds

사용 사례

01Breaking out of repetitive 'infinite loops' where the AI keeps trying the same failed solution.

02Pushing the AI to conduct deep research and documentation searches instead of relying on stale knowledge.

03Ensuring total ownership of complex debugging tasks that involve multiple files and dependencies.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tanweai/pua pua

For use in Claude.ai and ChatGPT