Server has 7 tools with significant definition gaps. Tool names lack action verbs (linux_ops, esxi_ops, router_ops, switch_ops, deco_ops, commvault_ops, windows_ops are all noun-based or generic). Descriptions vary widely: linux_ops, esxi_ops, router_ops, switch_ops, deco_ops are 50-150 chars (acceptable); commvault_ops is 600+ chars (verbose, token-wasteful); windows_ops is 150 chars. Most critically, parameter descriptions are minimal or absent. linux_ops and windows_ops accept 'command' and 'target_host' with only 1-2 sentence descriptions. esxi_ops, router_ops, switch_ops, deco_ops accept generic 'command' or 'action' params with no format guidance, enum constraints, or examples of valid values. commvault_ops is the exception, it has detailed parameter docs and multiple execution modes, but it's also the most complex and hardest to use correctly. No output schemas are documented for any tool. No error handling guidance. No pagination support despite tools that could return large result sets (commvault list actions). Security: linux_ops and windows_ops run arbitrary commands on all servers if target_host is omitted, dangerous default behavior with no confirmation step.

• Rename tools to verb_noun pattern: run_linux_command, execute_esxi_command, check_router_status, query_switch_ports, configure_deco_network, manage_commvault_backup, run_windows_command. This lets LLMs parse intent from the name alone.
• Add enum constraints to 'command' and 'action' parameters. For linux_ops, document valid command categories (system info, network, process, file, user). For switch_ops, enumerate valid actions (check_port_status, get_poe_info, list_vlans). For deco_ops, enumerate valid actions (list_devices, get_node_status, enable_guest_network). This prevents hallucinated commands.
• Document output schemas for all tools. Specify what fields each tool returns (e.g., linux_ops returns {status: string, stdout: string, stderr: string, exit_code: int}). Include field types and descriptions so LLMs can chain tools correctly.
• Add confirmation/dry-run support to linux_ops and windows_ops. Require a 'confirm=true' parameter for destructive commands, or offer a 'dry_run=true' mode that shows what would execute without running it. This prevents accidental infrastructure damage.
• Add error handling guidance to all tools. Document common failure modes (SSH timeout, authentication failure, command not found) and what the LLM should do (retry, check credentials, try a different command). Return structured error responses with actionable next steps.