The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests untrusted data from external Figma URLs via the get_design_context tool in SKILL.md. Malicious text within a Figma design (e.g., in layer names) could attempt to influence the agent's output. Evidence:
Ingestion points: get_design_context(fileKey, nodeId) in SKILL.md.
Boundary markers: Absent for raw Figma data.
Capability inventory: File-write access for generating .uxml, .uss, and .cs files.
Sanitization: Instructions require the agent to interpret data as a design spec rather than direct code transliteration, providing some isolation.
[EXTERNAL_DOWNLOADS]: The skill downloads image and SVG assets from URLs provided by the Figma MCP server into the local Unity project's Assets directory (documented in Step 5 of SKILL.md). This includes instructions to use localhost sources when provided by the MCP bridge. These downloads are core to the asset migration workflow.
[COMMAND_EXECUTION]: The skill generates and writes C# scripts and UI configuration files to the local file system. While these are not executed by the agent itself, they constitute the creation of executable logic based on external design inputs.

figma-to-unity