caira

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill dynamically fetches and reads public GitHub repository content (e.g., via the GitHub API endpoints like https://api.github.com/repos/microsoft/CAIRA/contents/... and the repo blob/download URLs described in SKILL.md and references/*) and uses that untrusted, public content (README.md, *.tf, etc.) to drive decisions and scaffold actions, which could allow indirect prompt injection from third-party files.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly performs runtime fetches from the CAIRA repository APIs (e.g., https://api.github.com/repos/microsoft/CAIRA/contents/reference_architectures?ref=) and uses returned README.md and *.tf content to directly drive agent prompts, recommendations, and scaffold generation, making that external content a required dependency that controls instructions.