flutter-security-expert
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted Dart source code (lib/**/*.dart), creating a surface for indirect prompt injection where instructions embedded in the target code could attempt to influence the agent's behavior.
- Ingestion points: Source code enters the agent context through the Read, Glob, and Grep tools during the audit process.
- Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' warnings when processing analyzed code.
- Capability inventory: The skill has access to potentially impactful tools including Bash, Write, and Edit.
- Sanitization: No specific sanitization or escaping of the ingested code content is described in the workflow.
Audit Metadata