workflow-deployment

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-supplied input as context for the deployment workflow without using boundary markers. This creates an indirect prompt injection surface where external instructions could influence high-privilege operations.\n
  • Ingestion points: User prompt passed to /workflow-start in SKILL.md.\n
  • Boundary markers: No delimiters or ignore-instructions warnings are used.\n
  • Capability inventory: Workflow includes Docker, Kubernetes, and CI/CD operations via commands like /code, /test, and /sre-review.\n
  • Sanitization: No sanitization or validation of user input is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 07:25 AM
Security Audit — agent-trust-hub — workflow-deployment