adb-screen-detection

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands via the adb CLI tool to perform screen captures and input actions. The implementation in scripts/adb-screen-capture.py and scripts/adb-tap-coordinate.py follows security best practices by passing arguments as a list to subprocess.run, effectively preventing shell injection.- [DYNAMIC_EXECUTION]: The script scripts/create-coordinate-maps.py contains a call to import('datetime'). This is used as a shorthand for generating a timestamp in a metadata dictionary and does not involve dynamic execution of untrusted code or loading from variable paths.- [INDIRECT_PROMPT_INJECTION]: The TOON workflow definitions (e.g., workflow/ocr-detection.toon) utilize parameters such as search_text which are interpolated into shell commands executed on the device (e.g., grep -i '{{ search_text }}'). While this represents a potential injection surface if the upstream orchestrator does not perform sanitization, it is a common pattern for such automation tools.
  • Ingestion points: search_text parameter in workflow/ocr-detection.toon and workflow/ocr-detection.md.
  • Boundary markers: None identified in the workflow definitions.
  • Capability inventory: The skill has the capability to execute shell commands on a connected Android device via the adb-shell action (as seen in perform_ocr phase).
  • Sanitization: Not present within the skill's workflow logic; sanitization would be expected at the orchestrator level.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 12:32 PM
Security Audit — agent-trust-hub — adb-screen-detection