k8s-argocd-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements strong data protection policies by explicitly instructing the agent never to access Kubernetes Secrets (e.g., argocd-secret) or repository credentials. It directs the agent to use ConfigMaps for configuration verification instead.
  • [SAFE]: All modification and destructive operations, such as kubectl patch, kubectl annotate, and kubectl delete, are strictly gated. The instructions require the agent to obtain explicit user confirmation ("Ask first") or provide the command for the user to execute manually.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from kubectl logs and resource statuses.
  • Ingestion points: Commands like kubectl logs and kubectl describe in SKILL.md and references/sync.md fetch external data into the agent context.
  • Boundary markers: The skill uses specific diagnostic patterns (e.g., table-based error mapping in references/comparison-errors.md) to guide interpretation.
  • Capability inventory: The agent has access to kubectl patch and kubectl delete (specified in SKILL.md and references/sync.md).
  • Sanitization: No explicit sanitization of log output is performed, but the requirement for human-in-the-loop confirmation for any subsequent actions effectively mitigates the risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 11:22 PM