The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to parse arbitrary document formats (PDF, Word, Excel, PPT) into structured Markdown for use in RAG systems. This represents a significant attack surface where instructions embedded in external documents could be processed by the agent and influence subsequent stages of a task pipeline.
Ingestion points: The skill takes a file_path as input, which points to local files that may contain untrusted data from external sources.
Boundary markers: The skill's instructions do not specify any delimiters or safety markers to wrap the parsed Markdown content, nor do they instruct the agent to ignore instructions embedded within the processed text.
Capability inventory: The skill script (scripts/run_doc_parse.py) performs file system operations, creating directories and writing output files (doc_parse.json).
Sanitization: There is no evidence of sanitization or filtering of the document content during the parsing process to neutralize potential prompt injections.
[COMMAND_EXECUTION]: The skill includes a standalone Python execution script (scripts/run_doc_parse.py) which is intended to be run by the agent. This script performs directory creation and writes parsed data to the local file system. While this is expected functionality for a document parser, it constitutes local command execution and file system interaction.

doc-parse