The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/setup.sh script downloads official MediaPipe face detection model weights (.tflite files) from storage.googleapis.com. This is a well-known service used for hosting official models and is considered a safe source.
[COMMAND_EXECUTION]: The skill instructs the user to execute a bash setup script and provides python3 command-line examples for frame extraction. These commands are standard for this type of developer-oriented tool.
[PROMPT_INJECTION]: The skill processes untrusted video data and extracts text using OCR (EasyOCR). This text is then stored in the manifest.json output file, which represents a surface for indirect prompt injection if downstream tools process this text without sanitization.
Ingestion points: The scripts/extract_clean_frames.py script reads and OCRs content from video frames provided by the user.
Boundary markers: The output JSON does not utilize specific boundary markers or warnings to encapsulate or flag the extracted text.
Capability inventory: The skill scripts have permissions to write to the file system (saving images and manifest files) and were initially installed via a setup script that executes shell commands.
Sanitization: No sanitization or filtering is performed on the text strings extracted from video frames before they are written to the manifest file.

avatar-frames