image-fallback-url-retry-in-composition
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a programming pattern that ingests external image URLs, which represents a surface for indirect prompt injection if the source data is compromised. * Ingestion points: The skill logic processes
record.imageUrlwithin Kotlin source code examples. * Boundary markers: There are no explicit delimiters or boundary markers mentioned for the ingested data. * Capability inventory: The pattern involves network image fetching via the AsyncImage component and UI state mutation. * Sanitization: The documented logic employs idempotent string replacement and conditional guards to mitigate the risk of infinite request loops during error processing.
Audit Metadata