drone-inspection-specialist
Warn
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
colmapbinary and external Python scripts such astrain.py,render.py, andconvert_to_web.pyusingsubprocess.runto perform 3D reconstruction tasks. These commands are constructed dynamically within theGaussianSplattingReconstructorandGaussianSplatInspectionToolclasses.\n- [DATA_EXFILTRATION]: TheFireAlertSystemclass includes asend_to_dispatchmethod that utilizesrequests.postto transmit alert payloads—including geolocation coordinates, timestamps, and temperature data—to a remote endpoint defined at runtime.\n- [PROMPT_INJECTION]: The skill usesmcp__firecrawl__firecrawl_searchandWebFetchto ingest data from external websites, creating a surface for indirect prompt injection. Ingestion points:firecrawl_search,WebFetchinSKILL.md. Boundary markers: None identified in the processing logic. Capability inventory:Bash,Write,Edit, and subprocess calls. Sanitization: No sanitization or filtering of retrieved web content is implemented before it enters the agent's context.\n- [EXTERNAL_DOWNLOADS]: The reconstruction pipeline generates an HTML viewer that fetches JavaScript libraries fromcdn.jsdelivr.net(Three.js). Additionally, the skill's reliance on theultralyticslibrary for YOLO models often leads to the automatic download of pre-trained model weights from official repositories.
Audit Metadata