The Agent Skills Directory

Dynamic Execution of Local Scripts: The snapo-network intercept command utilizes runpy.run_path to load and execute Python route handlers from a file path provided by the user. This mechanism allows the skill to perform complex HTTP response overrides by running arbitrary module-level code within the specified file.
Command Execution via Subprocess: The tool relies on the adb (Android Debug Bridge) command-line utility for communicating with connected devices. It resolves the adb executable from environment variables or user input and programmatically executes system commands to manage port forwarding, inspect sockets, and retrieve process metadata.
Data Sanitization (Redaction): To protect sensitive user information during network inspection, the skill includes a sanitization routine that redacts specific HTTP headers, such as Authorization, Cookie, and Set-Cookie, from its output.
Indirect Prompt Injection Surface: As a network inspector, the skill captures and displays potentially untrusted data from external network sources.
Ingestion points: Raw network request and response bodies are ingested in scripts/snapo-network through adb shell commands and local socket connections.
Boundary markers: The output is structured as NDJSON or plain text, but it lacks specific delimiters or warnings to instruct the agent to ignore potentially malicious embedded content.
Capability inventory: The implementation script has access to subprocess.run (for system interaction), runpy.run_path (for code execution), and socket operations.
Sanitization: While authentication headers are redacted, the content of request and response bodies is not sanitized for potential natural language instructions targeting the AI agent.

snap-o-network-inspector