The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied content to generate slide outlines and image prompts, creating a surface for indirect prompt injection.
Ingestion points: User-provided markdown content (e.g., source.md) enters the agent context during the analysis step.
Boundary markers: The references/base-prompt.md template uses headers like ## SLIDE CONTENT to separate instructions from data, but it lacks specific directives to the image model to ignore instructions potentially embedded within that user data.
Capability inventory: The skill utilizes scripts/gen_slide.py for network-based API calls (OpenAI and Gemini) and filesystem writes, while scripts/merge-to-pptx.ts and scripts/merge-to-pdf.ts create local documents.
Sanitization: No evidence of content sanitization or filtering is present in the scripts before the source material is used in prompt generation.
[COMMAND_EXECUTION]: The skill workflow involves the execution of several internal scripts, including gen_slide.py, overlay_logo.py, and Bun-based TypeScript scripts for merging slides into PPTX and PDF formats.
[EXTERNAL_DOWNLOADS]: The skill relies on third-party dependencies from standard registries, such as openai, google-genai, Pillow, python-dotenv, pdf-lib, and pptxgenjs, which are installed via pip and managed via npx.

q-presentations