devops-specialist
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The skill targets sensitive configuration files that define execution environments and automated pipelines, which constitutes a potential attack surface for indirect instructions. \n
- Ingestion points: Processes high-impact files including
.github/workflows/*.yml,Dockerfile,docker-compose.yml,vercel.json, andwrangler.toml. \n - Boundary markers: Absent. The instructions do not provide delimiters or specific warnings to ignore instructions embedded within the processed configuration files. \n
- Capability inventory: The skill enables the agent to write, optimize, and manage build and deployment configurations, which could be exploited to hijack CI/CD processes. \n
- Sanitization: Absent. There is no requirement for the agent to validate or sanitize contents before modifying deployment scripts or environment templates.
Audit Metadata