The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard shell utilities like grep, sed, head, and wc, along with the lit CLI tool to manage and query document data stored in temporary files.
[EXTERNAL_DOWNLOADS]: Instructions guide the user to install @llamaindex/liteparse via NPM and use the uv tool for managing Python dependencies (bm25s, aiofiles). These originate from well-known technology organizations and official registries.
[PROMPT_INJECTION]: The skill processes untrusted document data from files (PDF, DOCX, images) which introduces an indirect prompt injection surface.
Ingestion points: Document content extracted via lit parse in SKILL.md.
Boundary markers: None present to distinguish document content from system instructions.
Capability inventory: Shell execution (bash) and Python execution (search.py).
Sanitization: No sanitization of parsed document text before it is returned to the agent context.

effective-liteparse