adb-navigation-base

Pass

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes local Python scripts (adb-tap.py, adb-swipe.py, adb-wait-for.py) which execute ADB shell commands to interact with Android devices.
  • [EXTERNAL_DOWNLOADS]: The skill specifies dependencies on standard, well-known Python packages (pytesseract, pillow, numpy) for screen processing and character recognition.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it uses OCR to ingest text from the Android device's screen to drive automation logic.
  • Ingestion points: Text extracted from screen captures via pytesseract in adb-wait-for.py and action verification steps.
  • Boundary markers: None identified; the agent interprets OCR-derived text directly as UI state or success confirmation.
  • Capability inventory: Ability to perform UI interactions (taps, swipes) on the device and execute local Python orchestration scripts.
  • Sanitization: No sanitization or validation of the text extracted from the device screen is documented before it influences the agent's control flow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 17, 2026, 08:31 AM