The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses various shell commands for its core operations, including package management via pip3 and apt, image metadata inspection using the file command, and browser-based rendering through chromium.
[DYNAMIC_EXECUTION]: The skill frequently executes Python code snippets using python3 -c to perform tasks such as OCR extraction, image pre-processing (contrast/sharpness enhancement), and image format conversion.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process text extracted from user-provided images via OCR. This creates a surface where malicious instructions embedded in images could be ingested into the agent's context.
Ingestion points: User-uploaded images and URLs processed by vision_analyze, Tesseract, EasyOCR, and PaddleOCR.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the extracted OCR text.
Capability inventory: The skill has access to shell execution (python3, chromium), file system writes, and network operations (requests).
Sanitization: There is no evidence of sanitization or filtering of the text extracted from images before it is used in subsequent prompts.
[PRIVILEGE_ESCALATION]: The skill's setup instructions include commands that require elevated system privileges, specifically sudo apt install for Tesseract OCR and ssh root access for configuring remote proxy servers on Vultr.
[EXTERNAL_DOWNLOADS]: The skill guides the user to install multiple external dependencies, including pymupdf, Pillow, marker-pdf, easyocr, and paddleocr. It also interacts with various well-known AI services for image processing and generation.

image-recognition