The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill provides templates in references/c-harness.md for building test harnesses that execute binary code directly from files. These templates use platform-specific APIs like mmap with PROT_EXEC (Linux) or VirtualAlloc with PAGE_EXECUTE_READWRITE (Windows) to create executable memory regions and call them via function pointers.
[DYNAMIC_EXECUTION]: Python scripts in the skill utilize the unicorn engine for CPU emulation, allowing the execution and tracing of assembly code instructions within a virtualized environment.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze potentially untrusted data, such as assembly source files and binary shellcode. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the data being analyzed to influence the agent's logic.
Ingestion points: Raw binary data is read from local files (e.g., shellcode.bin, stub.bin) using open().read() and fread() across various Python and C scripts.
Boundary markers: The provided code templates do not include specific delimiters or warnings to prevent the agent from interpreting embedded instructions.
Capability inventory: The skill possesses capabilities for file system access, shell command execution (e.g., compilers, debuggers), and dynamic code execution via memory allocation.
Sanitization: There is no evidence of sanitization or validation of the binary data before analysis or emulation.
[COMMAND_EXECUTION]: The workflow requires the use of system-level tools such as gcc, nasm, gdb, objdump, and WinDbg, including commands that attach to running processes.

asm-testing