The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze text content directly from Figma nodes, which represents a surface for indirect prompt injection if the design content contains malicious instructions.
Ingestion points: The skill uses scan_text_nodes and get_node in SKILL.md to read arbitrary text from a Figma design.
Capability inventory: The agent has the ability to write back to the design using set_text, duplicate content with clone_node, and export visual data via get_screenshot.
Boundary markers: There are no instructions or delimiters defined to prevent the agent from following instructions that might be embedded within the text nodes it processes.
Sanitization: No sanitization or validation of the text content is mentioned before it is processed by the AI for structural analysis or adaptation.

text-replacement