daytona-integration

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). Most links (daytona docs and API, localhost) are normal documentation/API endpoints, but the direct downloadable shell script (https://claude.ai/install.sh) and the skill's recommendation to curl | bash make this set high-risk because executing remote .sh content can deliver malware even if hosted on a known domain.

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

• Third-party content exposure detected (high risk: 0.80). The skill executes and ingests content from public third‑party sources — for example installClaudeCLI runs curl -fsSL https://claude.ai/install.sh | bash and performs npm installs from the public registry into sandboxes — allowing untrusted external content to be introduced that the agent may subsequently observe or act on.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill includes an install step that runs curl -fsSL https://claude.ai/install.sh | bash at runtime (in installClaudeCLI), which fetches and directly executes remote shell code, making https://claude.ai/install.sh a high-risk runtime dependency.