iterating-with-ai-and-mcp
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: No security issues detected. The skill defines a standard set of Model Context Protocol (MCP) tools intended for a UI hot-reloading and visual verification loop within an IDE environment.\n- [PROMPT_INJECTION]: (Category 8) The skill defines a workflow for ingesting untrusted data (source code and device screenshots), which represents a surface for indirect prompt injection. 1. Ingestion points: Kotlin source files and device screenshots. 2. Boundary markers: Absent in the skill instructions. 3. Capability inventory: hotswan_reload, hotswan_build_and_install, hotswan_select_variant. 4. Sanitization: Absent; the skill relies on the agent's platform safety filters and the scope of the developer's local environment.
Audit Metadata