The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions that direct the AI model to bypass certain refusal behaviors. Specifically, the references/base-prompt.md file instructs the agent to create "stylistically similar alternatives" for sensitive or copyrighted figures rather than refusing to generate them. Additionally, the skill processes untrusted user-provided content and interpolates it into prompts for an external image generation tool, presenting a surface for indirect prompt injection.\n
Ingestion points: User-provided text is saved to source.md and read by the agent to generate slide outlines and prompts.\n
Boundary markers: The skill uses structured templates with specific tags like STYLE_INSTRUCTIONS and SLIDE CONTENT to separate system instructions from user content.\n
Capability inventory: The agent can write to the local file system, execute scripts via npx, and trigger external image generation skills.\n
Sanitization: No explicit content sanitization or safety filtering of the input text is documented before it is used to generate prompts.\n- [COMMAND_EXECUTION]: The skill executes local TypeScript scripts (scripts/merge-to-pptx.ts, scripts/merge-to-pdf.ts) using the npx -y bun command to assemble slide images into PDF and PowerPoint files. It also utilizes basic shell commands for file system operations such as checking for configuration files and managing output directories.\n- [EXTERNAL_DOWNLOADS]: The skill depends on third-party Node.js packages, specifically pdf-lib and pptxgenjs, which are used by the merging scripts to generate documents. These are well-established libraries in the technology ecosystem.

baoyu-slide-deck