The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The Python code example for AI integration demonstrates a pattern vulnerable to indirect prompt injection by interpolating untrusted input directly into the model context without safety boundaries.\n
Ingestion points: The analyze_with_claude(content: str) function in SKILL.md accepts external content for processing.\n
Boundary markers: The template lacks delimiters or instructions to ignore embedded commands within the content variable.\n
Capability inventory: The skill environment includes extensive project tools, such as gcloud for deployment, poetry for backend management, and npm for frontend operations, which could be targeted by a successful injection attack.\n
Sanitization: The provided code snippet does not include input validation or escaping mechanisms before passing data to the AI model.

cc-skill-project-guidelines-example