nano-tech-diagrams

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs staging files that produce presigned (signed) URLs and explicitly requires passing those presigned URLs verbatim into MCP tool calls and curl commands, causing the agent to handle and emit secret-like tokens in its output.

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

• Third-party content exposure detected (high risk: 0.80). The skill's File Staging section explicitly accepts arbitrary http(s) URLs ("When image_path is already an http:// or https:// URL (Fal media, any public host)") so the agent ingests images from public/untrusted hosts and those images are processed by tools (image_to_image, whiteboard_cleanup) that can materially influence subsequent decisions and actions.