agent-creator
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The architecture relies on interpolating untrusted data (tool results and user goals) into system prompts for evaluators and planners, which is a surface for indirect prompt injection.\n
- Ingestion points: Prompt templates in
references/prompts.mduse variables like{{output}},{{sources}}, and{{text}}to hold data from external sources.\n - Boundary markers: The templates do not use explicit delimiters or instructions to ignore embedded commands within the interpolated data.\n
- Capability inventory: The resulting agents have capabilities for tool execution and database persistence (TypeORM/PostgreSQL).\n
- Sanitization: The provided
renderPromptutility performs simple string replacement without escaping or sanitizing potential injection sequences.\n- [SAFE] (SAFE): No malicious code, hardcoded credentials, or obfuscation patterns were detected in the skill instructions.
Audit Metadata