The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill templates are designed to ingest untrusted user data such as dataset descriptions and goal statements, creating a surface for indirect prompt injection. Ingestion points: prompt_templates found in 'skill-report.json' (e.g., 'Choose a Chart', 'Create a Static Chart'). Boundary markers: Absent; user input is directly interpolated into prompts without delimiters or instructions to ignore embedded commands. Capability inventory: The skill contains no executable scripts, shell tools, or network operations; its capabilities are restricted to documentation and manual planning. Sanitization: No input validation or filtering of external content is specified.
[REMOTE_CODE_EXECUTION]: The documentation references standard Python visualization libraries including Matplotlib, Seaborn, Plotly, and Altair. These are recognized, well-known libraries and do not represent a security risk in this context.
[COMMAND_EXECUTION]: The Python code block in 'SKILL.md' is a static plotting example using Matplotlib with hardcoded arrays. It does not involve shell execution or user-controlled command injection.

data-visualization