PPT Generator Skill

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [Indirect Prompt Injection] (MEDIUM): The prompt templates (e.g., lego_style_prompt.md, clay_style_prompt.md) directly interpolate untrusted source material (【原始资料】) into complex instructions for the LLM.
  • Ingestion points: The 【原始资料】 placeholder in all files within the templates/ directory.
  • Boundary markers: Absent. There are no delimiters or system-level instructions to ignore embedded commands within the user-provided text.
  • Capability inventory: The system generates PPT files and provides instructions for external image generation tools.
  • Sanitization: Absent. The skill relies on the LLM to process the raw input without prior filtering.
  • [Data Exposure] (LOW): The script scripts/generate_styled_ppt.py reads file paths from a user-controllable JSON input (image_path).
  • Evidence: The create_slide function uses os.path.exists(image_path) and slide.shapes.add_picture(image_path, ...) without validating that the path resides within an allowed directory.
  • Risk: An attacker could potentially craft an input JSON that forces the script to include sensitive local images or files in the generated PPTX artifact.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 09:50 AM