test_driven_agent
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOW
Full Analysis
- [Prompt Injection] (SAFE): No instructions were found that attempt to bypass safety filters or override core agent behavior.- [Data Exposure & Exfiltration] (SAFE): No evidence of sensitive file access, hardcoded credentials, or unauthorized network requests.- [Obfuscation] (SAFE): Content is clear and readable with no hidden characters, encoding, or multi-layer obfuscation.- [Indirect Prompt Injection] (LOW): The skill directs the agent to interact with codebase contents (Ingestion: file reads) using high-privilege tools (Capability: Bash, Write). Boundary markers and sanitization are absent in this markdown protocol, but the skill reinforces structured testing logic which acts as a logical constraint for agent behavior.- [Dependency & Execution] (SAFE): Mentions standard 'npm test' in documentation as a verification step, but does not download or execute remote code or packages within the skill logic.
Audit Metadata