pptx-creator

Pass

Audited by Gen Agent Trust Hub on Sep 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied presentation data that can contain embedded instructions for image generation.\n
  • Ingestion points: Markdown outlines and JSON structure files are read by scripts/create_pptx.py.\n
  • Boundary markers: No specific delimiters or safety instructions are used to separate user data from control directives.\n
  • Capability inventory: The skill calls an external script via subprocess.run to handle generate: prompts.\n
  • Sanitization: The input string following the generate: keyword is not sanitized before being passed to the subprocess call, making it a potential vector for injection into the downstream tool.\n- [COMMAND_EXECUTION]: The skill executes external scripts using the subprocess module.\n
  • Evidence: scripts/create_pptx.py invokes uv run to execute an image generation script from the nano-banana-pro skill. While it uses argument lists to avoid shell injection, it executes external code based on user-influenced parameters.\n- [DYNAMIC_EXECUTION]: The skill dynamically constructs command-line arguments to execute Python scripts at runtime.\n
  • Evidence: The skill uses Path(__file__) to resolve relative paths to other scripts and builds a command list dynamically to generate images based on slide content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 29, 2026, 01:56 AM