The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as CSV files, tables, and spreadsheets to generate visual charts. This ingestion of external content creates an attack surface for indirect prompt injection if malicious instructions are embedded within the data.\n
Ingestion points: Data is read from files via Read, Glob, and Grep or searched via moda file search (as noted in SKILL.md).\n
Boundary markers: The instructions do not explicitly mandate the use of delimiters or specific 'ignore' warnings for the processed data.\n
Capability inventory: The skill can execute shell commands via Bash(moda:*), modify canvas content, and export files.\n
Sanitization: No specific sanitization or validation steps are provided for the raw data before it is passed to the charting tool.\n- [DYNAMIC_EXECUTION]: The skill utilizes a sandboxed JavaScript environment (moda canvas edit) to perform batch updates and mutations on the canvas. While this involves runtime code generation and execution, the environment is strictly constrained by the platform, prohibiting dangerous operations like eval(), require(), or direct network access.\n- [COMMAND_EXECUTION]: The skill makes extensive use of the moda CLI toolset to manage canvases, brand kits, and exports. This command execution is the intended primary interface for the skill's functionality and is restricted to the moda:* namespace.\n- [EXTERNAL_DOWNLOADS]: The documentation for brand kits describes using curl to download temporary previews of brand logos. This allows the AI agent to visually verify brand assets during the design process. These downloads are part of a documented, legitimate workflow for the vendor's platform.

moda-chart