The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute various third-party analysis tools such as exiftool, binwalk, steghide, zsteg, pngcheck, ffmpeg, and tshark to inspect files for hidden data.
[EXTERNAL_DOWNLOADS]: The instructions recommend installing domain-specific Python packages for steganography tasks, specifically stegano, sstv, and morse-audio-decoder via pip.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing untrusted media files (images, audio files, and PCAP network captures) which are potential surfaces for indirect prompt injection.
Ingestion points: Files are opened and processed using PIL.Image.open, wave.open, standard file open(), and tshark -r for PCAP analysis.
Boundary markers: No explicit boundary markers or instructions to ignore embedded instructions are provided when processing these files.
Capability inventory: The skill has access to Bash (shell execution), Write (file modification, specifically for PNG dimension correction), and WebFetch/WebSearch tools.
Sanitization: The provided Python scripts perform direct bit/byte extraction and data parsing without sanitization of the extracted content.

ctf-stego