The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Figma URLs, screenshots, and design tokens which could potentially contain malicious instructions meant to influence the agent's output.
Ingestion points: Figma API JSON responses, design token files, and image analysis (SKILL.md).
Boundary markers: None explicitly mentioned in the instructions for separating design content from system instructions.
Capability inventory: Network access via curl, file system writes for component generation, and MCP tool execution.
Sanitization: No explicit sanitization or validation of strings extracted from design layers or tokens.
[DATA_EXFILTRATION]: The skill uses curl to fetch design data from api.figma.com. This is a well-known service required for the skill's primary function, and the requests target official API endpoints.
[COMMAND_EXECUTION]: Provides example commands using curl and jq for data retrieval and processing. These are standard development tools used for their intended purpose in this context.
[CREDENTIALS_UNSAFE]: The skill correctly instructs users to store the Figma Personal Access Token in an environment variable (FIGMA_TOKEN) and warns against hardcoding it in generated code or commits.

figma-to-code