The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill requires administrative privileges to install necessary utilities and to perform memory scanning tasks on running processes.
Evidence: Use of sudo apt install scanmem and sudo scanmem in SKILL.md and references/memory-scanning.md.
[DATA_EXFILTRATION]: The skill provides scripts to access and modify the /proc/<pid>/mem interface, allowing direct interaction with a process's internal memory space, which can contain sensitive runtime information.
Evidence: Python scripts in SKILL.md and references/memory-scanning.md performing open(f'/proc/{pid}/mem', 'r+b', 0) and subsequent read/write operations.
[DYNAMIC_EXECUTION]: Python scripts for memory patching and network protocol analysis are constructed and executed dynamically using shell heredocs.
Evidence: `python3
<<'EOF'blocks found in multiple sections ofSKILL.md`.
[COMMAND_EXECUTION]: The skill utilizes a variety of system tools to analyze binaries and network traffic.
Evidence: Execution of checksec, readelf, ldd, and tcpdump for reconnaissance.
[EXTERNAL_DOWNLOADS]: The skill installs third-party software and libraries to facilitate game security analysis.
Evidence: Installation of scanmem via apt and blackboxprotobuf via pip3.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as game binaries and network traffic while possessing the capability to modify system files and process memory.
Ingestion points: Processes external binaries, save files, and network capture data.
Capability inventory: Can write to files and process memory, and execute shell commands.
Boundary markers: No markers are used to isolate or ignore instructions embedded in the processed data.
Sanitization: Data is processed directly without sanitization or validation of content.

game-technique