The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute a bundled Python script (scripts/audit_pptx_quality.py) to perform structural quality checks on the generated PowerPoint files. This script utilizes standard libraries (zipfile, xml.etree) and operates locally on the agent's output files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from research papers, PDFs, and scientific URLs which could contain malicious instructions.
Ingestion points: The skill extracts metadata, text, and figure legends from user-supplied papers (Step 1 of the workflow).
Boundary markers: While explicit prompt delimiters are not used, the skill implements a 'Terminology Ledger' to enforce technical consistency and minimize influence from external phrasing.
Capability inventory: The skill has the ability to write to the file system and execute its bundled audit script.
Sanitization: The skill relies on the python-pptx library for file generation, which provides structured XML handling.
[DYNAMIC_EXECUTION]: The skill generates a task-specific Python script (build_presentation.py) as a deliverable for the user. This script is intended for reproducibility of the PowerPoint deck and is generated based on the task context using standard templates.

nature-paper2ppt