The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes untrusted user data (topic descriptions and uploaded data files) which are interpolated into presentation content and chart generation scripts.
Ingestion points: User responses to the INTAKE.md questionnaire and user-uploaded CSV/JSON data files.
Boundary markers: The orchestration files do not specify explicit boundary markers or 'ignore' instructions for the data processing stages.
Capability inventory: The skill executes Python scripts (chartkit.py, generate_charts.py) via subprocesses and interacts with the document-skills:pptx subagent.
Sanitization: No explicit sanitization or validation of user-provided text is documented before it is used in script generation.
Dynamic Execution (LOW): The orchestration workflow involving generate_charts.py relies on generating and executing Python code at runtime. This behavior is considered low risk as it follows predefined templates specifically for data visualization and uses standard libraries like Matplotlib and Pandas.

ppt-creator