The Agent Skills Directory

[Indirect Prompt Injection Protection]: The skill processes external user documents (PDFs, notes, web pages) to generate study aids. It implements robust security measures by instructing the agent to treat all extracted strings as untrusted text and mandating the use of JSON serialization and character escaping. These practices are designed to prevent malicious content within the source material from executing scripts or breaking out of the data structure in the generated HTML application.
[Local Data Persistence]: The interactive study app uses the browser's localStorage to save a user's progress and spaced-repetition data. This is a standard functional feature for study applications, keeping data localized to the user's browser environment.
[Sandboxed Execution Guidance]: The instructions recommend validating the generated output in a sandboxed environment without allow-same-origin permissions, which is an effective defensive measure to ensure the generated study app remains isolated from sensitive browser contexts.

study-flashcards