The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install several Python packages to enable its functionality.
Evidence: pip install pillow reportlab, pip install pyobjc-framework-Vision pyobjc-framework-Quartz, pip install easyocr, and pip install pytesseract.
[COMMAND_EXECUTION]: The skill uses system command-line utilities and package managers to perform its tasks without requesting user confirmation.
Evidence: Uses sips for image conversion (sips -s format png <input> --out <output>), brew for Tesseract installation, and pip for Python package management.
[REMOTE_CODE_EXECUTION]: The skill dynamically generates and executes a Python script at runtime to produce the final PDF document.
Evidence: Instruction in SKILL.md to "Create a Python script using PIL and reportlab libraries" and "Execute the script to generate the PDF".
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes untrusted text from user-provided images via OCR.
Ingestion points: Image file paths provided by the user (SKILL.md steps 1-3).
Boundary markers: No specific delimiters or "ignore instructions" warnings are used to separate OCR-extracted content from the agent's internal logic.
Capability inventory: The skill has access to shell command execution (sips, pip, brew) and the ability to generate and execute Python code.
Sanitization: There is no evidence of sanitization, filtering, or validation of the text extracted from documents before it is passed to the translation or PDF generation steps.

visa-doc-translate