repomix-reference-black-tortoise

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWSAFE
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill is designed to ingest contents from 1,214 external source code files into the agent's context, creating a surface for indirect prompt injection.
  • Ingestion points: references/files.md contains the full text of the referenced codebase.
  • Boundary markers: Files are structured with ## File: <path> headers and line numbers to help the agent differentiate between content and instructions.
  • Capability inventory: The skill is passive and informational; it contains no definitions for command execution (eval, exec), file modification, or network operations.
  • Sanitization: No content filtering or sanitization of the source code is performed beyond excluding binary files and truncating base64 strings.
  • [Data Exposure] (SAFE): The skill configuration in references/summary.md demonstrates a strong security posture by explicitly ignoring sensitive paths (e.g., .env, .aws/credentials, .ssh/, .git/, and various lock files).
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 06:11 AM