skills/cosmix/loom/loom-argocd/Gen Agent Trust Hub

loom-argocd

Fail

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: HIGHDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches Kubernetes manifests from the Argo Project's official GitHub repository for installation.
  • [DATA_EXFILTRATION]: Contains instructions to access sensitive local file paths, specifically ~/.ssh/id_rsa, during repository registration.
  • [DATA_EXFILTRATION]: Includes commands to retrieve sensitive administrative credentials, such as the argocd-initial-admin-secret, from the Kubernetes environment.
  • [PROMPT_INJECTION]: Exhibits an indirect prompt injection surface through the ingestion of external Git repository data without boundary markers or sanitization, combined with shell execution capabilities. Evidence: 1. Ingestion points: Git repositories defined in repoURL fields within Application manifests. 2. Boundary markers: Absent. 3. Capability inventory: Bash (kubectl/argocd CLI) and Write. 4. Sanitization: Absent.
  • [COMMAND_EXECUTION]: Executes multiple shell commands using kubectl and argocd to manage cluster resources and application lifecycle.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 25, 2026, 07:35 AM
Security Audit — agent-trust-hub — loom-argocd