The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user-provided briefs and external project artifacts to generate the final slide deck structure.
Ingestion points: User-provided topic briefs and intent (Step 1) and external JSON artifacts like resolved_manifest.json (Step 2).
Boundary markers: The instructions do not define clear delimiters or use safety instructions to isolate ingested content from the slide-generation logic.
Capability inventory: The skill executes the agent-slides tool via uvx, reads multiple local project files, and writes JSON and PPTX files to the project directory.
Sanitization: No sanitization, validation, or escaping of the ingested data is described before the content is interpolated into slide operations.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes an external package at runtime using the uv package manager.
Evidence: The process relies on uvx --from agent-slides to access slide rendering and QA functionality. This tool is recognized as a vendor-specific resource for the author 'mpuig'.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to perform preflight checks, rendering, and quality assurance.
Evidence: Calls to slides preflight, slides render, and slides qa are issued via uvx throughout the project lifecycle.

slides-build