The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes structured component specifications which act as ingestion points for untrusted data. These specifications are rendered into multiple formats such as PDF, Email, and Web UIs, creating a surface where malicious content could influence the agent's behavior or lead to UI-based attacks. Ingestion points: spec objects passed to rendering functions in SKILL.md and rules files. Boundary markers: No delimiters or warnings are specified for content within the spec objects. Capability inventory: File system writing via renderToFile and network operations through email transport. Sanitization: The instructions do not define sanitization or escaping of external content before it is interpolated into the specifications.
[METADATA_POISONING]: The skill metadata lists the author as 'OrchestKit', which differs from the provided context author 'yonatangross'.

multi-surface-render