The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads local files and transmits their binary content to an external API endpoint determined by the UNSTRUCTURED_API_BASE_URL environment variable. While this is the intended purpose of the skill, it involves sending document data to a user-defined external service. The inclusion of an --insecure flag in the script allowing the bypass of SSL certificate verification further increases the risk of data exposure during transit if used in non-debug environments.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents an ingestion surface for indirect prompt injection.\n
Ingestion points: The script mineru_fulltext_extract.py reads content from local documents (PDF, Office files, images) supplied via the --file command-line argument.\n
Boundary markers: No explicit boundary markers or instructions are provided to the agent to treat the parsed document content as untrusted or to ignore embedded instructions.\n
Capability inventory: The skill script has capabilities for network communication (urllib.request.urlopen) and file system write access (Path.write_text).\n
Sanitization: The extraction process normalizes markdown underscore escapes (\\_ to _) but does not perform sanitization, filtering, or safety checks on the textual content extracted from documents before it is returned to the agent context.\n- [DYNAMIC_EXECUTION]: The test suite scripts/tests/test_mineru_fulltext_extract.py utilizes the importlib module to dynamically load the main script from the file system for unit testing purposes.

document-granular-decompose