PentQuiver exposes 5 read-only tools with clear verb-noun naming (search_commands, list_sources, get_command, get_arguments, resolve_command). All tools have descriptions (10 - 100 chars, within baseline 34 - 392 range). Input schemas are visible and properly typed. However, parameter descriptions are minimal or absent (e.g., 'values' in resolve_command lacks guidance on format/constraints), and output schemas are not documented. Error handling converts exceptions to ValueError but provides minimal recovery guidance. No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Composition is sound, tools chain logically (search → get → resolve). Overall: functional but lacks LLM-optimization depth.

• Document output schemas for all tools. E.g., search_commands returns list[{id: str, name: str, command: str, description: str, tags: list[str], source: str}]. Include this in docstrings or a separate schema registry.
• Enhance parameter descriptions with format/constraint details. For 'values' in resolve_command, add: 'Dictionary of {placeholder_name: value_string}. All values must be strings; they are substituted into the command template.'
• Add tool annotations via FastMCP. Decorate all tools with @mcp.tool(readOnlyHint=True, idempotentHint=True) to signal safety to agents.
• Improve error messages with recovery hints. Catch CommandNotFound and return: 'Command "<id>" not found. Available sources: [list]. Try search_commands(query="...") to find similar commands.'
• Add pagination guidance to search_commands description: 'Results are limited to 25 by default. Use limit parameter (max 1000) to retrieve more. For large result sets, consider refining the query.'
• Document the command_id format in get_command, get_arguments, resolve_command descriptions. E.g., 'Format: source/command_name (e.g., "nmap/nmap-aggressive-scan"). Call list_sources() to see available sources.'