The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves ingesting user-supplied descriptions and text to generate HTML posters and complex visual schematics. This creates a surface for indirect prompt injection where adversarial input could potentially influence the agent's execution environment or the resulting document output.
Ingestion points: User-provided content for poster sections and natural language descriptions used to generate AI visual elements via the scripts/generate_schematic.py tool.
Boundary markers: The instructions lack explicit boundary markers or warnings to the agent regarding the handling of untrusted user input within the HTML template or command-line arguments.
Capability inventory: The skill has access to Bash for command execution (including browser and office tool invocation) and Write/Edit for file generation.
Sanitization: There are no instructions for sanitizing or escaping user-provided content before it is interpolated into the HTML template or passed as arguments to shell-based scripts.
[COMMAND_EXECUTION]: The skill instructions involve constructing and executing shell commands through the Bash tool to generate graphics and convert document formats. Specifically, it uses python, google-chrome (headless), and libreoffice. The pattern of passing user-provided descriptive strings directly into these command-line calls increases the risk of command injection if the agent fails to properly escape the input.

pptx-posters