The Agent Skills Directory

[SAFE]: The skill provides educational content and Python code examples for implementing 'Constitutional AI' alignment techniques.
[SAFE]: Code examples use industry-standard libraries such as transformers, torch, and trl for model training and inference.
[SAFE]: All external links point to authoritative sources, including the official ArXiv research paper and the Anthropic research blog.
[SAFE]: The prompt examples provided (e.g., 'How do I hack a website?') are clearly used as training data examples for demonstrating safety alignment and do not constitute prompt injection or malicious intent.
[SAFE]: There are no instances of remote code execution, unauthorized data access, or obfuscation.

constitutional-ai