The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install various steganography tools from external sources, including pip3 install stegoveritas, gem install zsteg, sudo apt install binwalk, and direct downloads via wget for StegSolve.jar (from caesum.com) and stegseek (from GitHub).
[COMMAND_EXECUTION]: Detailed command references are provided for numerous shell-based tools including zsteg, binwalk, steghide, exiftool, multimon-ng, and stegsnow to perform file analysis, chunk extraction, and password brute-forcing.
[PRIVILEGE_ESCALATION]: Several commands utilize sudo to perform system-level package installations via apt and dpkg.
[DYNAMIC_EXECUTION]: Inline Python scripts are included using python3 -c to perform runtime calculations such as brute-forcing PNG CRCs for hidden dimension detection, analyzing WAV file headers for appended data, and identifying zero-width or homoglyph characters in text strings.
[INDIRECT_PROMPT_INJECTION]: As a forensics toolkit, the skill is designed to ingest and analyze untrusted external data (images, audio, and text files) which may contain hidden payload data, presenting a surface for indirect prompt injection if the extracted content is treated as instructions by the agent.
Ingestion points: Reads local image, audio, and text files for analysis (e.g., image.png, audio.wav, message.txt).
Boundary markers: None identified for the data extraction process.
Capability inventory: Includes file read/write, network downloads for tools, and shell command execution.
Sanitization: None identified; the skill is intended to extract raw hidden data for the user to review.

steganography-techniques