The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/chain_server.py functions as a Command & Control (C2) delivery server. It implements sophisticated User-Agent and JavaScript feature gating to selectively serve exploit stages (renderer exploits, sandbox escapes) to specific target builds while serving benign content to others to evade detection.
[PERSISTENCE]: The script scripts/electron_audit.py includes a snapshot_gadget function designed to perform binary patching of the v8_context_snapshot.bin file. This allows an attacker to inject malicious JavaScript into a frequently used builtin, ensuring persistence within every isolate of a targeted Electron application even if ASAR integrity fuses are enabled.
[DYNAMIC_EXECUTION]: The skill employs multiple dynamic execution techniques. scripts/sandbox_escape.js dynamically instantiates WebAssembly modules to allocate RWX (Read-Write-Execute) memory regions in the renderer process for shellcode injection. scripts/electron_audit.py facilitates runtime patching of V8 heap snapshots.
[COMMAND_EXECUTION]: The scripts/d8_debug.sh utility executes the V8 shell (d8) with specialized flags such as --allow-natives-syntax and --expose-gc specifically to facilitate the development and execution of memory corruption exploits.
[METADATA_POISONING]: The skill metadata explicitly identifies its purpose for Mitre ATT&CK techniques such as T1203 (Exploitation for Client Execution) and T1055 (Process Injection), confirming its role as an offensive weaponization toolkit.

browser-exploitation