The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like 7zz, asar, pkgutil, and xar to extract files from application archives (DMG, AppImage, pkg) and asar bundles. These operations are essential to its function of analyzing bundled JavaScript code.
[EXTERNAL_DOWNLOADS]: Instructions are provided to install several well-known libraries and tools from the NPM registry, including webcrack, @babel/parser, @babel/traverse, @babel/generator, and @babel/types. These are standard utilities for JavaScript deobfuscation and transformation.
[SAFE]: The skill implements a robust deobfuscation strategy using Abstract Syntax Tree (AST) manipulation via the Babel library. This is a secure alternative to dangerous practices like using eval() or exec() on untrusted input code.
[PROMPT_INJECTION]: An indirect prompt injection surface is present as the skill is designed to ingest and process external, untrusted JavaScript files for deobfuscation.
Ingestion points: The target JavaScript files are read as input by the deobfuscate.js and rename-locals.js scripts.
Boundary markers: There are no explicit markers or warnings in the pipeline to ignore instructions embedded in the target code (e.g., within comments or strings).
Capability inventory: The skill possesses file read/write capabilities (fs.readFileSync, fs.writeFileSync) and uses external extraction tools.
Sanitization: Input code is parsed using @babel/parser, which safely processes the code into an AST but does not filter for potentially malicious natural language instructions that might influence the agent after analysis.

re-js-app