Agent Skills Hub

context-engineering-collection

A structured repository of Agent Skills for context engineering, multi-agent architectures, and production-grade agent system optimization.

scientific-critical-thinking

Evaluate scientific claims and research methodology for rigor, bias, and validity. Use evidence-based frameworks like GRADE and Cochrane to analyze experiments, protocols, and study conclusions.

code-formatter

Universal CLI tool to convert and synchronize AI agent skills between Claude Code and Gemini CLI extensions.

identify-assumptions-existing

Stress-test existing product feature ideas by identifying risky assumptions across Value, Usability, Viability, and Feasibility using a multi-perspective devil's advocate framework.

presentation-reviewer

AI-powered PowerPoint presentation reviewer using top-tier consulting standards (McKinsey, BCG) for structure, storytelling, clarity, design, and DSFR compliance.

multi-llm-advisor

Fetches expert perspectives from OpenAI Codex and Google Gemini for architecture, code reviews, and debugging, with transparent LLM synthesis.

prompt-improver

Enriches vague prompts by performing codebase research and asking targeted questions to clarify user intent before execution.

inno-prepare-resources

Automates research resource preparation by loading instances, searching GitHub for codebases, building dataset descriptions, and downloading arXiv papers.

Verification & Quality Assurance

A robust verification and QA system for software agents featuring real-time truth scoring, automated code validation, and instant rollback capabilities to maintain high reliability.

prompt-optimizer

Transforms vague or poorly structured prompts into optimized, high-performance instructions using proven prompt engineering principles for better AI model execution.

security-hardening

Comprehensive security audit and hardening for AI agents: credential scanning, PII protection, prompt injection defense, and workspace config optimization.

novelty-check

Verify research idea novelty against recent literature. Use when user says '查新', 'novelty check', or needs to confirm if a method is original.