The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Figma files, such as review comments and design node content, which could theoretically contain malicious instructions.
Ingestion points: Data is ingested via the scripts/figma_rest.py helper script and MCP tools like get_metadata and get_design_context (SKILL.md, references/rest-comments.md).
Boundary markers: The instructions do not mandate the use of specific delimiters or warnings to ignore embedded instructions within Figma data.
Capability inventory: The skill has the capability to perform network operations, write to the file system via scripts, and execute shell commands for host configuration.
Sanitization: No specific sanitization or filtering of the content retrieved from Figma is mentioned in the instructions.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to install, configure, and authenticate the Figma MCP server on various host applications.
Evidence: Commands such as codex mcp add, claude plugin install, and cursor-agent mcp login are used for environmental setup (references/host-verification.md).
Context: These commands are part of the intended administrative functionality for server discovery and authentication.
[EXTERNAL_DOWNLOADS]: The skill references and downloads official configuration and plugin code from Figma's managed infrastructure.
Evidence: Fetches plugin information from github.com/figma/mcp-server-guide.git and interacts with the official remote server at mcp.figma.com (references/host-verification.md, SKILL.md).

figma-mcp