loom-research

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill mandates preserving and outputting raw source wording verbatim (Raw Capture) when fetching content, so any API keys or secrets present in those sources would be reproduced by the model and thus exposed.

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly requires fetching and ingesting content from open sources—e.g., "Input (URLs, files, text, Material List)" and "Capture flow: Fetch source content using AI web tools" (including "social thread")—so the agent will read untrusted, user-generated public content that can materially influence its reading and synthesis workflow.