The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/JsEnv_Dev.js contains a WebSocket client that uses the eval() function within its _execjs handler to execute arbitrary JavaScript code received from a remote server. This mechanism allows for dynamic runtime execution of instructions within the browser context.
[DATA_EXPOSURE_AND_EXFILTRATION]: The JSRPC architecture (demonstrated in scripts/JsEnv_Dev.js and scripts/emit_flask_proxy.py) establishes a bidirectional communication channel between a target web browser and a local Flask proxy. This design is specifically intended to extract runtime data such as encryption keys, signatures, and tokens from a web page's memory to an external process.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process content from arbitrary target URLs provided by the user (as seen in SKILL.md and references/workflow-recon.md). This creates an attack surface where a malicious website could serve crafted content or anti-debugging scripts intended to manipulate the agent's logic or exfiltrate processed data.
[EXTERNAL_DOWNLOADS]: The documentation file references/antidebug/dynamic-alias.md includes a reference to an external GitHub repository (https://github.com/0xsdeo/Fuzz_Crypto_Algorithms) as a tool recommendation for fuzzing crypto parameters.

js-reverse-automation