video-debugger

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

• Third-party content exposure detected (high risk: 0.85). The skill's workflow (SKILL.md Phase 4.5 and the references like references/diff-tools.md) explicitly instructs downloading and using third‑party models from Hugging Face/LM Studio (e.g., huggingface-cli, https://lmstudio.ai) and to read the generated vision-analysis.txt outputs from those models, meaning the agent ingests and acts on untrusted third‑party model outputs that can influence subsequent analysis and actions.