The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate-slides.py utilizes subprocess.check_call to execute shell commands. It specifically runs pip install google-genai at runtime to ensure the Gemini API client is available. Executing arbitrary commands via shell wrappers is a high-risk capability that can be exploited if arguments are not strictly controlled.
[REMOTE_CODE_EXECUTION]: The skill's workflow involves the runtime installation of the google-genai Python library. While the library originates from a known provider, dynamic dependency management at execution time bypasses standard security vetting of the environment and introduces potential supply chain vulnerabilities.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external, untrusted documents to generate slides.
Ingestion points: Untrusted data enters the agent context through user-provided academic papers (PDFs) and Markdown files (e.g., source-paper.pdf, source.md).
Boundary markers: Absent. The instructions do not define delimiters or provide 'ignore embedded instructions' warnings for the content analysis phase.
Capability inventory: The skill has significant capabilities, including file system writes across multiple scripts and network access for image generation via the Gemini API.
Sanitization: Absent. Content extracted from the source documents is directly interpolated into prompts for the image generation model without escaping or validation.

paper-slide-deck