mirrord-temporal

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied parameters such as Kubernetes namespaces, workload names, and JQ filters.
  • Ingestion points: Parameters are gathered from the user for use in YAML generation and cluster discovery commands in SKILL.md.
  • Boundary markers: The instructions explicitly state to treat user-supplied values as data only and not as instructions.
  • Capability inventory: The skill utilizes kubectl for reading cluster state and troubleshooting as seen in SKILL.md.
  • Sanitization: The skill provides specific guidance to validate Kubernetes names against a regex and to reject shell metacharacters before interpolating them into commands.
  • [COMMAND_EXECUTION]: The skill instructs the agent to use kubectl commands for discovery and troubleshooting.
  • The commands are limited to read-only operations (e.g., kubectl get, kubectl config, kubectl cluster-info).
  • The instructions include explicit mitigations against command injection by validating user-provided identifiers before they are used in shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 09:06 AM
Security Audit — agent-trust-hub — mirrord-temporal