The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the native macOS utility sips within scripts/story/generate-images.py and scripts/story/verify-images.py to check image dimensions and aspect ratios. This is a benign use of a built-in system tool to ensure output quality.
[EXTERNAL_DOWNLOADS]: The skill downloads educational images from well-known and trusted sources such as Wikimedia Commons and US Federal Government archives (NASA, USGS, etc.) as documented in references/chapter-images-guide.md. These operations use standard Python libraries and target official repositories.
[REMOTE_CODE_EXECUTION]: The skill utilizes npx to execute the @marp-team/marp-cli tool for generating presentation slides. This is a standard industry practice for using well-known development tools and is restricted to the specific, reputable marp-team organization on the NPM registry.
[CREDENTIALS_SAFE]: The skill correctly handles sensitive information by requiring API keys for ElevenLabs and Google Gemini to be provided via environment variables (ELEVENLABS_API_KEY, GEMINI_API_KEY). It explicitly warns against hardcoding these keys, adhering to security best practices.
[DATA_INTEGRITY]: The references/verified-infographic-guide.md establishes a comprehensive 8-phase verification workflow. This process is designed to prevent the fabrication of data in infographics by requiring citations, performing web-based fact-checking, and conducting post-render audits to ensure visual output matches verified data.
[INDIRECT_PROMPT_INJECTION]: The skill ingests textbook content to generate media. While this presents an attack surface for indirect prompt injection, the skill mitigates this through structured layout specifications and "locked" prompts that prevent the image models from re-authoring or inventing data based on untrusted inputs.

book-media-generator