The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external Figma node layers to generate and update documentation files.
Ingestion points: The skill uses the get_node_info tool (Step 1 in SKILL.md) to fetch textual and structural data from external Figma URLs provided by the user.
Boundary markers: Instructions in SKILL.md advise the agent to use only Figma content and to notify the user of specific platform mentions, but there are no technical delimiters or explicit instructions for the agent to ignore potentially malicious embedded commands within the fetched Figma data.
Capability inventory: The agent has the capability to write and modify local MDX files in the docs/content/components/ directory (Step 2 in SKILL.md).
Sanitization: The skill lacks explicit sanitization or filtering logic to handle adversarial text embedded in Figma layers before it is transformed into MDX content.
[COMMAND_EXECUTION]: The documentation references the execution of local project-specific scripts using the bun runtime.
Evidence: references/patterns.md mentions running bun docs:images:og:generate (which requires ffmpeg) and bun docs:images:og:validate to manage Open Graph assets.
Context: These commands are documented as part of the standard asset management workflow within the development environment and are not used for malicious purposes.

seed-migrate-component-docs-from-figma