wiki

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

• Third-party content exposure detected (low risk: 0.10). The workflow ingests web articles and papers via web_extract and processes user-provided sources and pasted text, but these are user-directed search/fetch operations rather than monitored feeds or unprompted outsider communication channels.

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

• Potentially malicious external URL detected (high risk: 0.70). The npm installation command instructs the runtime to download and execute obsidian-headless from the public npm registry, presenting an unverified external runtime dependency.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill prompt instructs the agent to modify system service files and run a sudo command (sudo loginctl enable-linger $USER) to configure a systemd service.