The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses child_process.spawnSync in scripts/run-pipeline.js to coordinate the deobfuscation pipeline by running internal scripts on the target file.
[REMOTE_CODE_EXECUTION]: The deobfuscation engine (specifically ob-variant-pass.js, geetest4-guarded-pass.js, and tonghuashun-order-pass.js) uses Node.js's vm.runInNewContext to evaluate segments of the analyzed JavaScript code. Because the vm module is not a secure sandbox, executing untrusted code fragments presents a risk of host system compromise if an attacker crafts a script to escape the execution environment.
[PROMPT_INJECTION]: The skill is designed to process external, untrusted JavaScript code. If the deobfuscated output contains instructions that the AI agent interprets as commands, it could influence the agent's behavior.
Ingestion points: The skill reads external JavaScript files provided as command-line arguments to the run-pipeline.js script.
Boundary markers: The skill does not employ explicit boundary markers or warnings to separate the analyzed code content from the agent's operating context.
Capability inventory: The skill possesses capabilities to execute shell commands (spawnSync), read and write to the filesystem (fs), and execute dynamic JavaScript (vm).
Sanitization: No sanitization is performed on the deobfuscated code before it is output or processed by the agent.

ast-deobfuscation