tao-generate-image-grounding

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly instructs the agent to collect API keys from the user (and mentions storing them in config fields like vlm.gemini.api_key), which would require the LLM to accept and potentially emit secret values verbatim into configurations or commands — an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The skill ingests outsider-authored free text via data.input_jsonl (each line’s caption) when running the auto_label generate pipeline, and that caption text is fed into the VLM at Step 0/1 for expression extraction and grounding.