发现优质的 AI Agent 技能

**Unified security audit toolkit** combining OWASP Top 10 vulnerability scanning, dependency analysis, secret detection, SSL/TLS verification, AI Agent security checks, and automated security scoring.

> Tổng hợp insights từ Anthropic's Skill System, OpenAI's GPTs Architecture, > Google DeepMind's Agent Design, và kinh nghiệm xây dựng hàng trăm production skills. > > "Skill tốt không phải skill viết hay — mà là skill hoạt động đúng 100 lần liên tiếp."

> **Biến AI Agent từ "chỉ biết nói" thành "biết LÀM" — chạy lệnh, thao tác hệ thống, > xử lý dữ liệu phức tạp.**

Evaluate expectations against an execution transcript and outputs.

Compare two outputs WITHOUT knowing which skill produced them.

Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.

You are the Rubric QA Agent. Score the implementation against all 220 criteria below. Inspect the actual code in the repository AND the forge workspace files (implementation_plan.json, spec.md, progress.log). Answer YES or NO for each. No partial credit.

You are the QA Agent in the FlowForge pipeline. Your job: review the completed implementation against the original spec and score it.

You are the Planner Agent in the FlowForge pipeline. Your job: turn a spec into a structured `implementation_plan.json`.

You are the Coder Agent in the FlowForge pipeline. Your job: execute every subtask in the implementation plan, in order.

Compact element references that reduce context usage dramatically for AI agents.

Complete reference for all agent-browser commands. For quick start and common patterns, see SKILL.md.