The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes command-line tools like curl to fetch external JavaScript files and web resources. This is a core part of its functionality to download target SDKs and bytecode for local analysis.
[REMOTE_CODE_EXECUTION]: The skill is designed to execute untrusted JavaScript code within a Node.js environment. It uses require and the vm module to load and run scripts. Additionally, it implements runtime hooks for eval and the Function constructor to intercept and modify code behavior (e.g., stripping anti-debugging logic) before execution.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external, potentially attacker-controlled JavaScript. Instructions or patterns within the analyzed code could be reflected in the diagnostic reports generated by the engine, potentially influencing the agent's logic during the environment patching process.
Ingestion points: Analyzed JavaScript files located in the source/ directory and the diagnostic reports generated by env_core.js in scripts/env_core.js.
Boundary markers: No explicit delimiters or instructions to ignore content within the diagnostic reports were found.
Capability inventory: The skill possesses file system access, dynamic execution capabilities via vm and require, and the ability to perform network requests via curl.
Sanitization: There is no evidence of sanitization or filtering of the diagnostic logs before they are presented to the agent for analysis.

env-patch