data-engineering-best-practices
Pass
Audited by Gen Agent Trust Hub on Feb 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious instructions, jailbreak attempts, or prompt injection patterns were identified. The operating modes and instructions are focused on technical data engineering advice.
- [SAFE]: No hardcoded credentials, sensitive file paths, or unauthorized data exfiltration mechanisms were found. The playbooks correctly promote secure practices like using Cloud Secret Manager for connection handling.
- [SAFE]: No remote code execution, obfuscation, or dynamic execution patterns are present. All guidance is delivered via static markdown files and structured templates.
- [SAFE]: No privilege escalation or persistence mechanisms were detected. The skill operates entirely within the context of the agent's interaction.
- [SAFE]: The skill processes user-provided PR diffs and DAG code, which represents an indirect prompt injection surface. Evidence: 1. Ingestion points: PR_REVIEW and AIRFLOW modes in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Generates markdown reports and ASCII diagrams; no direct tool-calling, file-system writing, or network execution capabilities. 4. Sanitization: Absent. This surface is inherent to the review functionality and is not exploitable given the agent's limited capabilities.
Audit Metadata