The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted binary data from game artifacts, assets, and network captures.
Ingestion points: Game binaries (.exe, ELF, .so, .dll), Unity asset files (.assets, .dmp), and network PCAP files.
Boundary markers: No explicit markers or delimiters are used to isolate content extracted from untrusted artifacts.
Capability inventory: File system write operations (open(..., 'wb')), network socket communication (socket.socket()), and direct process memory access via /proc/<pid>/mem.
Sanitization: Extracted data, such as string literals or metadata, is used directly in analysis logic without sanitization.
[COMMAND_EXECUTION]: The skill includes instructions for high-impact system operations necessary for reverse engineering.
Binary Patching: Provides Python snippets to modify executable logic, such as overwriting instructions at specific offsets to bypass win conditions.
Memory Manipulation: Contains instructions for reading and writing to process memory via the Linux proc filesystem and using memory scanning tools.
[DYNAMIC_EXECUTION]: The skill generates and executes logic at runtime using shell heredocs.
Evidence: Multiple instances of `python3
<<'EOF'` are utilized to execute dynamically generated scripts for process memory modification, binary patching, and network protocol interaction.

game-ctf