The Agent Skills Directory

Dynamic Execution: The skill provides a feature to intercept and override API responses using custom Python route handlers. It uses runpy.run_path to dynamically load and execute arbitrary code from a user-provided file path. While this is the core functionality for mocking network traffic, executing arbitrary scripts from the file system is a significant capability that warrants review.
Indirect Prompt Injection: The skill ingests untrusted network traffic data from an Android device to provide inspection logs. This creates a potential attack surface where maliciously crafted data within the network traffic could attempt to influence an AI agent analyzing the tool's output.
Ingestion points: The tool reads network events and data (CDP-style records) from Android devices via ADB abstract sockets in scripts/snapo-network.
Boundary markers: The tool redacts sensitive headers like Authorization and Cookie to limit data exposure, though it does not provide explicit boundary warnings to downstream agents.
Capability inventory: The script can execute adb commands, manage network forwards, and load/execute local Python scripts.
Sanitization: The script includes a redact_headers function to sanitize sensitive authentication and session data before emitting output.
Command Execution: The CLI tool communicates with Android devices by executing adb commands using subprocess.run. This includes listing devices, querying process information, and managing port forwarding, which are standard operations for Android development tooling.

snap-o-network-inspector