n-version-workflow

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): The skill implements an orchestration pattern where an agent processes data (code) generated by other sub-agents. • Ingestion points: Step 3 and Step 4 involve analyzing and testing code generated in Step 2. • Boundary markers: The instructions do not specify explicit delimiters or warnings for the analysis phase. • Capability inventory: The workflow involves spawning subprocesses and running automated tests. • Sanitization: No explicit sanitization or security-filtering steps are defined for the generated code before it is passed to the tester or reviewer agents.
  • Command Execution (SAFE): The skill references spawning subprocesses and running tests. Within the context of a development-focused agent skill, these are intended functional capabilities for orchestration. No suspicious shell commands, unauthorized file access, or hardcoded malicious payloads are present.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:58 PM