The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to invoke the libreoffice system binary for converting legacy .ppt files to .pptx and for rendering slides to PDF. While it uses argument lists to mitigate shell injection, it executes these commands using user-provided file paths as input.
[COMMAND_EXECUTION]: The skill executes an external Python script located at a hardcoded absolute path (/path/to/skills/sn-da-image-caption/scripts/caption.py) to process slide images. This dependency on an external script execution increases the attack surface.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it extracts content from untrusted PowerPoint files and incorporates it into the agent's context without sanitization or boundary markers.
Ingestion points: Text is extracted from slide shapes, tables, and chart titles in extract_all_slides_text, and captions are generated from slide images in caption_image_slides.
Boundary markers: Absent; extracted text is aggregated into a single string without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can execute system commands via subprocess.run for file conversion and image analysis.
Sanitization: No filtering or escaping is applied to the extracted text before it is presented to the agent.

ppt-analysis