iblai-vibe-local-llm

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

• Third-party content exposure detected (low risk: 0.10). The runtime path is the Tauri-command/event integration for local model download/status/chat (e.g., useModelDownload()/localLLMProps invokes download_model(modelId) and streams ollama_chat_stream, then listen() subscribes to model:download-progress and ollama:chunk/done/error), but this contract text does not ingest arbitrary outsider-authored free-form data without first selecting a specific model/catalog entry, and the only free text it handles is user-driven chat content rather than monitored external feeds/queues.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's troubleshooting notes and the install_ollama command describe fetching and executing the remote installer via "curl ... | sh" (explicitly referencing https://ollama.com/install.sh), which is a runtime fetch-and-execute of remote code.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly defines an install command (install_ollama) and notes the official installer writes to /usr/local/bin and creates a systemd service and user via sudo (i.e. modifies system files and users), which pushes an agent toward privileged system changes unless handled as a detection-and-guide flow.