cli

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The list includes an external ZIP install URL (https://example.com/skill.zip) referenced by the skill installer which is an untrusted archive download that could deliver malware; the remaining URLs are local dev endpoints, placeholders, or an official GitHub repo referenced in the docs and are lower risk.

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

• Third-party content exposure detected (medium risk: 0.30). In lh agent run, the user-supplied --prompt <text> is sent to the backend and then streamed back for execution, meaning outsider-authored free text is directly ingested by the agent runtime workflow.

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

• Potentially malicious external URL detected (high risk: 1.00). The lh skill install command fetches remote skill packages at runtime (examples: https://github.com/lobehub/skill-repo and https://example.com/skill.zip), and those fetched files contain skill content (prompts/instructions) that directly control agent behavior.