The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill has a high-risk attack surface for processing malicious instructions embedded in design data.
- Ingestion points: Ingests external design metadata and code connect snippets via figma.get_node and figma.get_code_connect_map.
- Boundary markers: Absent. The skill lacks instructions to delimit Figma content or warn the agent against following instructions found within it.
- Capability inventory: The agent is authorized to create and write .tsx and .css files in the project structure (e.g., src/design-system/).
- Sanitization: Absent. There are no guidelines or logic for escaping or validating the content retrieved from Figma before it is written to the filesystem.
Prompt Injection (SAFE): The skill uses authoritative language to enforce design fidelity, but no malicious pattern for bypassing safety filters or overriding core instructions was found.
Data Exposure (SAFE): No hardcoded credentials, API keys, or access to sensitive local file paths (like SSH or AWS configs) were detected.
Dependencies & RCE (SAFE): The skill does not perform external downloads, remote script execution, or unverifiable package installations.

atomic-design