The Agent Skills Directory

SAFE (SAFE): The skill package is composed exclusively of Markdown documentation, rules, and metadata. No executable code or scripts were detected in the 120 files.- CREDENTIALS_UNSAFE (SAFE): All code examples for server and client implementations (e.g., x402, MCP) use placeholders such as '0xYourAddress', 'API_KEY', or environment variable lookups like 'os.getenv("PRIVATE_KEY")' instead of hardcoding secrets.- EXTERNAL_DOWNLOADS (SAFE): Referenced URLs point to legitimate, official documentation and repositories for established protocols (e.g., Anthropic, Google, Coinbase). No malicious download-and-execute patterns are present.- PROMPT_INJECTION (SAFE): The instructions provided to the agent are purely technical and focused on protocol adherence. There are no attempts to override safety filters or hijack agent behavior.

ai