The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill supports a data-driven design workflow that processes external CSV data to generate graphic variants, creating a potential attack surface.
Ingestion points: The dataset.import command, as documented in references/recipes.json and references/upstream-0.12.md, accepts raw CSV strings or file paths to populate variables in vector documents.
Boundary markers: The instructions do not provide explicit delimiters or "ignore instructions" warnings for the agent when processing user-provided CSV data.
Capability inventory: The skill leverages the run_command tool to execute vector engine commands and file.save to write output files to the local file system.
Sanitization: There are no documented procedures for sanitizing or validating the contents of the CSV data before it is processed by the agent or the underlying engine.
[EXTERNAL_DOWNLOADS]: The documentation references external resources for obtaining the necessary command-line tools and for design guidance.
The skill instructs users or agents to verify the presence of vectorcraft-cli and provides a link to https://github.com/storytold/vectorcraft/releases for manual installation if the tool is missing.
Multiple external repositories are cited as sources for workflow logic, including github.com/creold/illustrator-scripts, github.com/ie3jp/illustrator-mcp-server, and the skillhub.cn platform.
These references are provided for installation guidance and workflow facts rather than automated remote code execution.

vectorcraft-scenario-playbooks