The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted design data which creates a potential surface for indirect prompt injection. * Ingestion points: The skill reads layer names (s.name) and text content (s.characters) from the design file via scripts/collectAccessibilityData.js. * Boundary markers: The skill instructions do not explicitly require the agent to use delimiters or ignore embedded instructions within these data fields when generating the final report. * Capability inventory: The skill uses the execute_code tool for data extraction and export_shape for design evidence gathering. * Sanitization: No sanitization is performed on the text strings extracted from the design before they are processed by the agent.

penpot-audit-accessibility