trellis3d

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly asks the user to paste their Pixazo API key into the conversation, instructs saving and using it (checking the conversation, writing ~/.pixazo/api-key, and including it in request headers), which requires the model to receive and potentially emit the secret verbatim in requests or outputs.

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime path includes fetching https://www.pixazo.ai/models/trellis3d.md (and possibly the HTML variant) to load the “full reference,” which is public web content authored by an outsider and can be ingested as readable text into the agent’s LLM context.