Improve resource tool description with comprehensive spec

- Add detailed overview of pinned vs indexed resources
- Clear when to use / when not to use guidelines
- Explain storage types and their purposes
- Add practical notes for better LLM usage
- Match quality of semantic_search tool spec

Author: kennvene-aws

crates/chat-cli/src/cli/chat/tools/tool_index.json

@@ -234,7 +234,7 @@
   },
   "resource": {
     "name": "resource",
-    "description": "PRIORITY TOOL: Unified resource management for both pinned (always included) and indexed (retrieved on demand) resources. Use this tool to manage context across chat sessions with two storage types: 'pinned' resources are always included in conversations, while 'indexed' resources are retrieved on demand via semantic search. Always use this tool first when you need to remember, recall, or retrieve any information from previous conversations.",
+    "description": "A unified resource management tool for storing and retrieving information across chat sessions.\n\n## Overview\nThis tool manages two types of resources:\n- **Pinned resources**: Always included in conversations (session context)\n- **Indexed resources**: Retrieved on demand via semantic search (knowledge base)\n\n## When to use\n- When you need to remember information from previous conversations\n- When looking for specific content the user has shared before\n- When the user asks about their stored resources or context\n- When you need to search through previously indexed content\n- When managing user's persistent knowledge base\n\n## When not to use\n- For general web searches or external information\n- When the user hasn't stored any relevant resources yet\n- For real-time data that changes frequently\n\n## Storage Types\n- **pinned**: Always included in every conversation (use for frequently needed info)\n- **indexed**: Retrieved via semantic search when relevant (use for large amounts of content)\n\n## Notes\n- Always use this tool FIRST when you need to recall previous conversations or user preferences\n- For search queries, use the user's exact wording to capture contextual cues\n- Break down complex searches into core concepts for better semantic matching\n- Use 'show' to see what resources are available before searching\n- Pinned resources appear automatically, indexed resources require explicit search",
     "input_schema": {
       "type": "object",
       "properties": {