The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of adb shell to execute diagnostic commands on a connected Android device. This includes system-level queries such as dumpsys, getprop, and top to monitor thermal state, battery usage, and process activity.
[DATA_EXFILTRATION]: The skill facilitates the collection of highly sensitive device data, specifically through the adb bugreport command. As noted in the documentation, these reports can contain account identifiers, app activity, network details, and notifications. However, the skill provides explicit instructions to redact sensitive information and requires informed user approval before any data is transmitted.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests and processes output from external adb commands, which could theoretically contain malicious data from a compromised device.
Ingestion points: Data enters the agent context through the output of adb shell commands and adb bugreport extractions.
Boundary markers: The skill lacks formal delimiters or LLM-specific boundary markers for command output, though it enforces a specific output format for the agent's findings.
Capability inventory: The agent has the capability to execute shell commands via the adb tool across multiple scripts and workflow stages.
Sanitization: The skill mandates manual redaction of sensitive fields (serials, phone numbers, accounts) by the agent/user before data sharing.

diagnose-android-overheating