The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes a pattern for modal expansion that ingests external data without proper sanitization.
Ingestion points: The openModalFromElement function in examples/shared-elements.md accepts a modalContent string as an argument.
Boundary markers: The content is interpolated directly into a template literal without delimiters or instructions to treat the data as untrusted.
Capability inventory: The skill demonstrates the ability to perform DOM updates and append new elements to the document.body via the updateFn callback.
Sanitization: No HTML escaping or sanitization logic is present in the example code before rendering the content.
[DYNAMIC_EXECUTION]: A code example in the shared elements documentation utilizes dynamic HTML generation.
Evidence: The function openModalFromElement in examples/shared-elements.md uses modal.innerHTML to inject the modalContent variable into a new div element.
Context: While presented as a pattern for UI transitions, this method creates a vulnerability surface where executable code could be injected if the modalContent originates from an untrusted source.

web-animation-view-transitions