framework-to-mastra

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security risks were identified. The skill acts as a code generator and scaffolding utility for the Mastra AI framework.- [COMMAND_EXECUTION]: The skill includes Deno scripts (analyze-framework.ts, validate-conversion.ts) used during development to parse markdown files and verify project structure. These scripts perform routine local file system operations required for code generation.- [EXTERNAL_DOWNLOADS]: Project documentation and templates appropriately reference well-known services and official software repositories, including OpenAI, Anthropic, Docker, and the Mastra framework. These references are part of standard configuration and deployment workflows.- [PROMPT_INJECTION]: The skill provides templates for building agents that ingest external data. Analysis of the Indirect Prompt Injection surface shows: (1) Ingestion points include framework files and user-provided situation descriptions; (2) Boundary markers are used in prompt templates to separate user data from instructions; (3) Capabilities include LLM generation and tool execution as defined in the Mastra framework; (4) Sanitization is implemented through mandatory Zod schema validation for all inputs and outputs.- [DATA_EXFILTRATION]: No mechanisms for unauthorized data exfiltration were found. The skill correctly instructs users to manage sensitive credentials like API keys using environment variables and cloud provider secret management tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 09:45 AM