The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The scripts scripts/batch-remove-background.sh and scripts/remove-background.sh contain logic to automatically execute npm install -g rmbg-cli if the required binary is missing. This installs unverified third-party code globally on the host system without an explicit user prompt or confirmation during script execution.
COMMAND_EXECUTION (LOW): The skill utilizes subprocess calls to execute powerful media tools including FFmpeg and ImageMagick. While the Python implementation in scripts/media_convert.py uses the safer list-argument method for subprocess.run, the skill provides a broad surface for processing potentially malicious media files using system binaries.
PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection through processed media content or metadata. Ingestion points: Media files and directory paths processed by shell and Python scripts. Boundary markers: None present to delimit untrusted data. Capability inventory: Subprocess execution and global package installation capabilities. Sanitization: No validation or sanitization of input filenames or media metadata is performed before processing.
DATA_EXFILTRATION (SAFE): No hardcoded credentials, sensitive file access (e.g., SSH keys), or unauthorized network exfiltration patterns were detected.

media-processing