alogin-admin — Administrator Skill

This skill is for human operators only. disable-model-invocation: true prevents Claude from selecting this skill automatically. It must be invoked explicitly with /alogin-admin.

Do not include this skill in an AI agent's system prompt or skill list. Commands here (approve, vault get, policy set) would allow an agent to bypass its own safety controls.

Use this skill for:

• Reviewing and acting on pending HITL approval requests
• Managing trust windows (time-bounded auto-approve)
• Reading or rotating vault credentials
• Configuring agent policies and server system prompts
• Monitoring the audit trail
• Validating and dry-running policies