The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of local Python scripts using the uv run tool and interacts with Android devices via adb shell commands to verify root status and module activity.
[PROMPT_INJECTION]: A surface for indirect prompt injection exists because the skill monitors adb logcat output to verify bypass effectiveness. Malicious applications on the connected device could write crafted strings to the log buffer to influence the agent's assessment of the security state.
Ingestion points: Monitoring device logs via adb logcat as described in the troubleshooting section.
Boundary markers: None identified; the documentation suggests simple keyword filtering (e.g., grep) without sanitization.
Capability inventory: Execution of Python scripts (preflight-validation.py) and potential triggering of downstream app automation (adb-karrot) based on validation results.
Sanitization: Not documented; the skill appears to rely on raw log output matching.

adb-bypass