The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes user-supplied text to generate document layouts.
Ingestion points: Raw text data enters the agent context via the text parameter in frame.create commands (found in references/recipes.json and references/workflows.md) and through external file reading via the file.place command described in references/command-evidence.json.
Boundary markers: The skill uses structured JSON parameters for all tool invocations and a specific reference syntax ($ref) to manage data flow, which provides logical delimitation of content.
Capability inventory: The skill utilizes tools for file system writing (file.exportPdf, file.exportPptx, file.saveAs) and document manipulation via the DesignCraft CLI or MCP service.
Sanitization: No explicit sanitization of user-provided strings is mentioned in the instructions, as the skill relies on the underlying design engine to handle text properties safely within the document structure.
[DYNAMIC_EXECUTION]: The skill includes a local Python utility (scripts/write_script.py) intended to be run by the user or agent to compile layout recipes into command sequences. This script performs logical remapping of data references and generates JSON output based on provided templates without utilizing dangerous functions like eval() or exec().

designcraft-scenario-playbooks