The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content from source files to generate slide outlines and image generation prompts. This constitutes an indirect prompt injection surface as malicious instructions in the source text could theoretically influence the downstream image generation process. However, given the output is restricted to raster images and the tool utilizes existing agent guardrails for image generation, the risk is minimal.
Ingestion points: The skill reads source content in Step 1.2 to create source.md and subsequent outlines.
Boundary markers: The skill uses structured templates and a multi-step confirmation workflow that includes user review of the outline and prompts before image generation, which serves as a significant manual mitigation.
Capability inventory: The skill invokes image generation tools and executes local scripts for file merging.
Sanitization: No specific sanitization of input text is described, relying instead on the user's manual review steps.
[COMMAND_EXECUTION]: The skill executes local TypeScript utility scripts (scripts/merge-to-pptx.ts and scripts/merge-to-pdf.ts) using the bun runtime to perform the final merging of generated images into presentation documents. These scripts are part of the skill's distribution and perform standard file-processing tasks using well-known libraries.

baoyu-slide-deck