The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the DesignSync MCP tool to manage design projects and assets. It follows a strictly defined execution sequence (list -> finalize_plan -> write_files) to ensure operational integrity and user oversight through plan confirmation.
[DATA_EXPOSURE]: The skill accesses local design assets, such as DESIGN.md, tokens, and brand logos, for the purpose of synchronizing them with Claude Design. It explicitly instructs the agent and user to anonymize sensitive business or customer data before proceeding with the upload.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data files, which represents a potential attack surface for indirect instructions.
Ingestion points: Reads DESIGN.md, token files (JSON/CSS), and image assets from local staging directories.
Boundary markers: The skill uses instructional guidance to define which files should be uploaded, though it lacks technical delimiters for the file content itself.
Capability inventory: Employs the DesignSync tool for plan finalization and file writing to a remote service.
Sanitization: Relies on manual user anonymization and licensing checks rather than automated technical filtering of file contents.

design-sync-upload