The Agent Skills Directory

Privilege Escalation (HIGH): The installation instructions include the use of sudo apt-get install, which requires administrative privileges. Granting an agent the ability to execute commands with sudo is a significant security risk.
Indirect Prompt Injection (HIGH): The skill processes untrusted external data (video, audio, and image files) using powerful CLI tools (FFmpeg, ImageMagick, RMBG).
Ingestion points: Files provided via -i (FFmpeg), magick [input] (ImageMagick), and rmbg [input] (SKILL.md).
Boundary markers: None identified. There is no instruction to the agent to ignore or sanitize metadata or embedded content in the files.
Capability inventory: The skill uses subprocess calls to execute complex media processing pipelines which can involve file creation and modification.
Sanitization: None. The skill assumes input files are safe, making it vulnerable to exploitation if a file contains malicious metadata or triggers vulnerabilities in the underlying media libraries (e.g., ImageTragick).
Unverifiable Dependencies (MEDIUM): The skill requires npm install -g rmbg-cli. This package is downloaded from the public NPM registry and does not come from a trusted source defined in the security policy, posing a supply-chain risk.
Command Execution (LOW): The skill rely on executing arbitrary shell commands for its core functionality. While this is the stated purpose, it increases the attack surface if the agent is tricked into modifying these commands.

media-processing