The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches Tesseract OCR language data from the tesseract-ocr/tessdata_best GitHub repository. This is a well-known service providing official resources for the Tesseract engine. This download is required for OCR functionality and is treated as a safe reference to a well-known provider.
[COMMAND_EXECUTION]: The skill operates via a CLI tool (lit) and interacts with local environment tools like LibreOffice to convert Office documents into PDFs. These interactions are standard for the skill's documented functionality and do not involve suspicious privilege escalation or persistence mechanisms.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from various document formats (PDF, DOCX, etc.) which are ingested via the parse method or CLI. Ingestion points include document parsing and directory batching (e.g., in scripts/batch_parse_dir.py). While no specific boundary markers or sanitization processes are explicitly enforced by the parser to prevent instructions within these documents from affecting agent logic, this is an inherent attack surface for any document processing tool. The environment capabilities (including Bash and file system access) are documented for the intended use of the parser.
[SAFE]: The skill demonstrates secure practices by instructing users to manage sensitive credentials for remote OCR services via environment variables rather than hardcoding them in the skill or command-line arguments. All external resources, including the liteparse Python and Node.js packages, are legitimate vendor resources provided by the author.

liteparse