The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted text content from user-provided images (e.g., passports, bank statements), creating a vector for indirect prompt injection.
Ingestion points: The skill ingests user-provided images via the /visa-doc-translate command as specified in SKILL.md.
Boundary markers: There are no instructions for the agent to use delimiters or ignore instructions potentially embedded within the OCR-extracted text.
Capability inventory: The skill possesses significant capabilities including shell command execution (pip, brew, sips), file system writing, and Python script execution.
Sanitization: The instructions do not define any sanitization or validation procedures for the text extracted from documents before it is translated or placed into the final PDF.
[COMMAND_EXECUTION]: The skill automatically executes shell commands and package managers to prepare its environment.
Evidence: SKILL.md instructs the agent to run pip install for multiple libraries and brew install tesseract. It also uses the native macOS sips utility for image conversion.
[REMOTE_CODE_EXECUTION]: The skill utilizes runtime script generation and execution to produce output.
Evidence: The instructions in SKILL.md require the agent to "Create a Python script using PIL and reportlab libraries" and subsequently "Execute the script to generate the PDF."

visa-doc-translate