The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides comprehensive instructions and JavaScript templates for building exploitation primitives such as addrof and fakeobj. These patterns are designed to achieve arbitrary memory read/write access by exploiting V8 engine bugs like JIT type confusion and incorrect bounds elimination, leading to full renderer process compromise.
[PRIVILEGE_ESCALATION]: The playbook contains a dedicated section on Chrome sandbox escape techniques, including a decision tree for transitioning from a compromised renderer to a full browser process compromise. It specifically identifies Mojo IPC interfaces as high-value targets for sandbox escape.
[DYNAMIC_EXECUTION]: The instructions detail advanced methods for dynamic code injection, such as identifying and overwriting WebAssembly (WASM) RWX pages with shellcode and using JIT spray techniques to bypass memory protections like constant blinding and W^X enforcement.
[COMMAND_EXECUTION]: Provides specific instructions for executing V8 with advanced debugging flags and natives syntax, such as --allow-natives-syntax and intrinsics like %OptimizeFunctionOnNextCall, which are used to bypass standard engine constraints during exploit development.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions for cloning the V8 source code from official Google repositories for the purpose of building custom, debuggable versions of the engine for exploitation research.

browser-exploitation-v8