---
name: agnesai-labs-ops-runbook
description: "Execute a reviewed runbook with approval and receipt verification."
license: "Apache-2.0"
metadata:
  decimalai-skill-id: "9c80e9f5-c33f-427b-9058-d87e09c51c25"
  decimalai-version-id: "9ea4349e-ae30-4a2f-ad07-08bed18f3316"
  decimalai-version: "1"
  decimalai-body-sha256: "5bcd9bf98dee465ea75bfd0143ef10f3248091cb4022cf4824e17907e885f643"
  decimalai-name: "agnesai-labs/ops-runbook"
  decimalai-format: "1"
  decimalai-invocation: "model"
  decimalai-category: "devops"
  decimalai-source: "https://app.decimal.ai/s/agnesai-labs-ops-runbook@1/SKILL.md"
---
# ops-runbook

Read the runbook and collect diagnostics first. Use only fixed author-approved argv through ToolContext.exec; the deployment selects the sandbox provider. A restart requires human approval even in the simulator. Verify the receipt. Stop after a failure or unknown result and investigate; do not blindly retry. Customer adaptation: replace argv with a bounded service allowlist and select an appropriate sandbox provider before startup.

Use official present for generated deliverables under fde-output/ops-runbook/. Ask the business choice with official ask_user_question and wait for its validated answer before the action; keep backend tool authorization separate. If official Plan mode is active, submit the fixed workflow with exit_plan_mode and wait for its native approval before business steps. Plan approval never replaces a later business choice or tool permission.