The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill's core function is to ingest untrusted data (PDFs) and have the agent interpret that data to generate code (HTML). This creates a significant injection surface.
Ingestion points: The agent reads slide screenshots (slide-NN.jpg) generated from user-provided PDF files.
Boundary markers: No delimiters or explicit instructions are provided to help the agent distinguish between slide content and potential malicious instructions embedded in the slides.
Capability inventory: The skill allows the agent to execute a Python script (pdf_to_images.py) and write multiple HTML files to the local filesystem.
Sanitization: There is no sanitization of the text extracted from the images. The instructions explicitly tell the agent to 'Match text content exactly,' which is exactly how an injection would be triggered.
[Unverifiable Dependencies] (LOW): The skill requires the installation of poppler via Homebrew. While poppler is a trusted utility, the skill relies on the user's environment having the correct binary available, and the Python script executes this binary via subprocess.run.
[Command Execution] (LOW): The script pdf_to_images.py uses subprocess.run to call pdftoppm. It correctly uses a list for arguments rather than a shell string, which mitigates direct shell injection risks.

slidekit-templ