guardian-cli-ai-pentest

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The list mostly contains harmless example/placeholder domains (example.com, testsite.example.com, api.example.com) and an innocuous vendor domain (ara.so), but it also includes a GitHub repository (github.com/zakirkun/guardian-cli.git) from an unknown user that the instructions tell you to clone and pip install—making the set potentially suspicious because unknown GitHub-hosted code installed locally can be used to distribute malware.

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

• Third-party content exposure detected (high risk: 0.75). Guardian’s runtime workflow ingests tool outputs (e.g., HTTP discovery/probing and vulnerability scanner results) as readable text into the LLM context; those tool outputs can include arbitrary free-form content fetched from outsider-controlled web targets/URLs (public pages, responses, scraped content), enabling indirect prompt injection.