The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Word documents containing slide titles and rasterized images, creating a potential vector for indirect prompt injection.\n
Ingestion points: The scripts/extract_deck.py utility parses anchor text and slide images from user-provided .docx files.\n
Boundary markers: The instructions lack specific prompt delimiters to distinguish extracted data from agent instructions.\n
Capability inventory: The skill is capable of reading and writing local files, including Excel worklists and commented Word documents.\n
Sanitization: No explicit sanitization or filtering is performed on the extracted text before it is presented to the LLM.\n- [COMMAND_EXECUTION]: The test suite scripts/test_extract_deck.py uses subprocess.run to call the project's own CLI script extract_deck.py for verification. This is a controlled execution of internal code.\n- [DYNAMIC_EXECUTION]: The scripts/extract_deck.py script dynamically modifies the Python path (sys.path) during its bootstrap process to locate internal vendor libraries. This behavior is restricted to discovery within a specific folder structure.

localise