init-agent-config

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security risks were identified during the analysis.
  • [PROMPT_INJECTION]: The skill contains no instructions that attempt to override safety protocols or jailbreak the agent. Directional language used (e.g., 'CRITICAL', 'IMPORTANT') is appropriately context-bound to configuration quality and repository safety.
  • [CREDENTIALS_UNSAFE]: The skill explicitly instructs agents to perform a cross-check to ensure 'no secrets or credentials' are included in the generated configuration files.
  • [REMOTE_CODE_EXECUTION]: All command patterns provided for auditing are standard local operations (e.g., cat, grep, jq) used to inspect project manifests and do not involve executing untrusted remote code.
  • [DATA_EXFILTRATION]: There are no patterns suggesting unauthorized network activity or data exfiltration.
  • [SAFE]: The skill addresses the potential risk of indirect prompt injection by instructing the agent to preserve existing boundary rules and verify all facts against the repository's code.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 10:38 AM