The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides code examples that introduce vulnerability surfaces, specifically an Open Redirect in API routes.
Ingestion points: In examples/api-routes.md, the GET handler for /api/redirect retrieves a url parameter directly from url.searchParams.get("url").
Boundary markers: There are no delimiters or instructions to validate the embedded URL string against a whitelist.
Capability inventory: The redirect(307, target) function is called using the unvalidated input, which allows the application to redirect users to arbitrary and potentially malicious external domains.
Sanitization: No URL validation or domain sanitization is present in the example code, facilitating open redirection attacks.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates the use of the {@html} tag for rendering content, which creates a significant surface for Cross-Site Scripting (XSS).
Ingestion points: In examples/core.md, the +page.svelte component renders data.post.content which is sourced from a database.
Boundary markers: None present for the rendered data.
Capability inventory: The {@html} tag renders raw HTML strings directly into the document object model, which can execute any embedded <script> tags or handle malicious event attributes.
Sanitization: The example does not demonstrate the use of a sanitization library to filter the database content before it is rendered, making the application vulnerable if the data source is ever compromised or contains user-provided content.

web-meta-framework-sveltekit