deploy-ops

Pass

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes platform-specific deployment commands, including kubectl apply, fly deploy, helm upgrade, and eb deploy. These commands are standard CLI operations for cloud and container orchestration services.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting user-provided data for deployment identifiers.
  • Ingestion points: User-supplied version numbers, PR numbers, and deployment tags (SKILL.md).
  • Boundary markers: The skill uses markdown code blocks to delimit command execution steps and tables for status reporting.
  • Capability inventory: Shell command execution via CLI tools (kubectl, fly, helm, eb) for deployment and monitoring (SKILL.md).
  • Sanitization: No explicit validation or escaping of user-provided version strings is defined within the instruction files.
  • [SAFE]: The skill incorporates defensive rules that mitigate operational risk, such as requiring all CI tests to pass, mandatory staging deployment before production, and the documentation of a rollback plan. It does not attempt to access sensitive local files or perform unauthorized network exfiltration.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 5, 2026, 08:06 PM
Security Audit — agent-trust-hub — deploy-ops