The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download the ghidra_flirt tool from a third-party GitHub repository (https://github.com/nneonneo/ghidra_flirt). While common in the reverse engineering community, this source is not part of the pre-defined trusted vendor list.
[REMOTE_CODE_EXECUTION]: The installation process for the external ghidra_flirt tool involves executing make immediately after cloning, which runs code from the downloaded repository.
[COMMAND_EXECUTION]: The skill makes extensive use of shell-based binary analysis tools such as objdump, readelf, otool, nm, and rz-bin to inspect user-provided files.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill is designed to ingest and parse untrusted binary files (PE, ELF, Mach-O) provided as samples.
Boundary markers: There are no explicit delimiters or warnings instructed for the data extracted from these samples.
Capability inventory: The skill uses shell command execution, Python script execution (pefile), and binary parsing capabilities.
Sanitization: The output from the binary analysis tools is interpolated into the agent's context without explicit sanitization or filtering for embedded instructions.

re-imports