THE TOOL REGISTRY

PRACTICE LAB CONNECTION In Chapters 65 to 67 you grow the Resume Review Agent's tool registry. Extend it with typed handlers, so file tools share path checks, and add version and deprecation fields. Test the prompts below against the registry you build. ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS [START OF PROMPT] Review this tool registry implementation for missing production-readiness fields (permission metadata, version, deprecation status, rate-limit config, dispatch/handler reference) and for any entries that are inconsistent with each other: [PASTE LAB tool_registry.py HERE]. Output format: a gap table (Tool, Missing Field, Why It Matters) plus a corrected registry entry example. [END OF PROMPT] [START OF PROMPT] Our agent is connected to [NUMBER] tools across [N] MCP servers, but only a handful are relevant per task. Design an on-demand tool-discovery method for this registry: [PASTE LAB OR PRODUCTION REGISTRY CONFIG HERE] that selects relevant tools per task, and estimate the token savings versus loading all tool definitions every turn for a typical task like [TASK EXAMPLE]. Output format: discovery-method design, then a before/after token estimate table. [END OF PROMPT]