datocms-frontend-integrations
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and rendering of content from external DatoCMS APIs, which is a standard surface for indirect prompt injection.
- Ingestion points: Content enters the application via
executeQuery,useQuerySubscription, and framework-specific data loaders (e.g.,performQueryin SvelteKit) as documented in the framework-specific reference files. - Boundary markers: The skill explicitly instructs the use of
data-datocms-content-link-boundaryon complex components like Structured Text blocks and inline records to isolate independent editing targets. - Capability inventory: The integration performs network requests to DatoCMS GraphQL and Real-time Update (SSE) endpoints. It does not perform arbitrary command execution or unrestricted file system writes.
- Sanitization: The skill follows best practices by providing a
stripStega()utility to clean metadata from strings before programmatic use (comparisons, SEO, etc.) and implements a robustisRelativeUrl()validator for user-supplied redirects to prevent open redirect attacks. - [PROMPT_INJECTION]: The deterministic flags for prompt injection are false positives. The terms 'override' and 'bypass' in the reference documentation describe standard architectural patterns for custom component rendering and cache management (e.g., 'bypass shared caching for draft requests'), rather than attempts to subvert agent safety policies.
Audit Metadata