The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.Popen in scripts/manage_services.py to start local services. It performs an automated search for binaries named jsrpc (and platform-specific variations) in common user directories like ~/jsrpc, ~/jstest, and ~/hacker/jstest. Executing unverified binaries from these locations represents a command execution risk if a malicious binary is placed in a searched path.
[DYNAMIC_EXECUTION]: The JSRPC environment script scripts/JsEnv_Dev.js implements a remote execution capability using eval(). It processes messages received via WebSocket to execute arbitrary JavaScript within the browser context. This is a powerful dynamic execution primitive characteristic of JSRPC-based reverse engineering tools.
[DYNAMIC_EXECUTION]: The skill generates Python and JavaScript files at runtime, such as flask_proxy.py and jsrpc_inject.js. These generated artifacts are then executed by management scripts using the local system's interpreters, which constitutes dynamic code generation and execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (JavaScript bundles, network traffic, DOM elements) to automate code generation, creating an attack surface for indirect prompt injection. If a target website contains malicious content specifically designed to subvert the analysis, it could potentially influence the code produced by scripts like detect_encryption.py or the diagnostic behavior of ast_candidate_analyzer.js.
Ingestion points: Data is ingested through browser interaction and network monitoring via the chrome-devtools-mcp interface.
Boundary markers: Standard formatting is used to separate data from instructions, but there are no specific defenses against adversarial data content.
Capability inventory: The skill manages a Flask proxy capable of network communication and a browser injection environment with full JavaScript execution capabilities.
Sanitization: Basic redaction of sensitive keys (password, token, etc.) and truncation of large values are performed in common.py and the runtime probe, providing some level of data safety.

js-reverse-automation