The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/gemini_image_pack.py accesses sensitive file paths by searching for and reading .env files in multiple locations, including the user's profile directory (~/.agents/skills/md-to-modern-pptx/.env). The retrieved API keys are then transmitted to a remote URL defined in the configuration.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to download image content from external URLs returned by the API gateway and saves them to the local disk.
[COMMAND_EXECUTION]: The skill requires the installation of external software packages pptxgenjs and markitdown (the latter from a trusted source) using package managers.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface when processing untrusted Markdown files for image generation prompts.
Ingestion points: The input Markdown file is read and parsed in scripts/gemini_image_pack.py and scripts/md-deepresearch-to-pptx.js.
Boundary markers: No boundary markers or instructions to ignore embedded commands are included in the prompt template used for image generation.
Capability inventory: The skill possesses file system write permissions for saving generated decks and images, and network connectivity via the image generation API.
Sanitization: No input sanitization or validation is performed on the Markdown content before it is interpolated into API prompts beyond basic Markdown symbol removal.

md-to-modern-pptx