The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/build_context.py utilizes subprocess.run to execute Python components and accepts a --repo argument that enables loading and executing code from an external directory path provided by the user.\n- [COMMAND_EXECUTION]: The runtime/tools/android_shell_utils.py script performs shell operations using adb, including attempts to execute commands with su to obtain elevated privileges on an Android device to access restricted memory maps.\n- [PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it ingests and summarizes content from external Android logs and previous analysis reports, which could contain adversarial text aimed at influencing agent behavior.\n- [DATA_EXFILTRATION]: The skill processes highly sensitive Android artifacts like memory heap dumps (HPROF) and system logs. While it provides privacy-focused summaries for the agent context, the management of these artifacts poses a data exposure risk if they are moved outside of authorized environments.

android-memory-evidence