luxonis-build-poc

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md requires cloning and reading outsider-authored public web content at runtime (e.g., https://github.com/luxonis/oak-examples and live docs starting from https://docs.luxonis.com/llms.txt), which would be ingested as readable text into the agent’s LLM context via the “Knowledge sourcing” and “Navigate the checkout” steps.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs the agent at runtime to fetch live docs from https://docs.luxonis.com/llms.txt and to clone https://github.com/luxonis/oak-examples (used as the implementation reference and to run/adapt example code), so external content is fetched during runtime and directly controls prompts and executed code.