flutter-patterns
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing and filtering conversation history and LLM responses, creating a surface where malicious data could influence application logic.\n
- Ingestion points: LLM response processing and history building logic described in SKILL.md (Section 12) and references/llm-prompt-isolation.md.\n
- Boundary markers: The skill emphasizes prompt isolation and history filtering but does not provide explicit delimiter-based boundary markers for untrusted content within the provided code examples.\n
- Capability inventory: The skill patterns involve state mutation via Riverpod, navigation via GoRouter, and file system operations like
_storage.deleteFilein SKILL.md.\n - Sanitization: An InputSanitizers utility class is provided in SKILL.md (Section 6) to handle specific digit sanitization and directionality checks.\n- [SAFE]: The integration of GenUI/A2UI is described as a declarative JSON-based widget composition method. The documentation explicitly clarifies that it does not involve runtime code execution or eval(), mitigating dynamic execution risks.\n- [SAFE]: All external documentation links and package references target well-known technology services and official Flutter/Google domains.
Audit Metadata