oci-compute

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

• Third-party content exposure detected (high risk: 0.85). OCI Functions section runs curl -LSs https://raw.githubusercontent.com/.../install | sh, which fetches outsider-authored free text (a GitHub-hosted script) at runtime and feeds it into the execution/LLM context path via the shell pipeline.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill includes "curl -LSs https://raw.githubusercontent.com/fnproject/cli/master/install | sh", which fetches and immediately executes a remote script at runtime to install the Fn CLI, satisfying the criteria for a runtime external dependency that executes remote code.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The prompt's cloud-init explicitly creates a deploy user with NOPASSWD sudo (creating a new account with root privileges) and writes to /etc/nginx/conf.d and runs systemctl/firewall-cmd, all of which modify system files/services and grant elevated access to the machine.