@gabrywu/knowledge-bank 0.1.2-alpha.276 → 0.1.2-alpha.312

package/README.md

@@ -12,6 +12,57 @@ Knowledge Bank is a lightweight Claude Code plugin that transforms your developm
 - 🏗️ **Intelligent organization** using multi-dimensional classification
 - ⚡ **Zero-friction integration** with your existing Claude Code workflow
 
+## Design Philosophy
+
+Knowledge Bank is built on a fundamental principle: **knowledge should be deeply integrated into the entire software development lifecycle**, not treated as a separate documentation burden.
+
+### 🔄 **Lifecycle-Integrated Knowledge Management**
+
+**Throughout Every Development Phase:**
+- **Design Phase**: Capture architectural decisions and design patterns
+- **Development Phase**: Record code patterns, implementation insights, and technical decisions
+- **Code Review Phase**: Document review findings, best practices, and improvement patterns
+- **Testing Phase**: Store testing strategies, edge cases, and quality insights
+- **Deployment Phase**: Record deployment patterns, configuration knowledge, and operational insights
+- **Maintenance Phase**: Capture troubleshooting knowledge, performance optimizations, and evolution patterns
+
+### 🔗 **Code-Bound Knowledge Strategy**
+
+**Knowledge-Code Coupling**: By binding knowledge collection and retrieval directly to your code workflow, we ensure:
+
+- **High Knowledge Density**: Every captured insight is contextually relevant and immediately actionable
+- **Automatic Freshness**: Knowledge stays current as it's updated alongside code changes
+- **Zero Maintenance Overhead**: No separate documentation process to maintain
+- **Contextual Relevance**: Knowledge appears exactly when and where you need it
+
+### 🤖 **LLM-Native Knowledge Design**
+
+**Implicit Information for AI Consumption**: Knowledge Bank recognizes that modern development increasingly involves AI assistance. Our knowledge format is optimized for:
+
+**LLM-Optimized Storage**:
+- Structured data formats that AI models can efficiently process
+- Rich metadata and classification for intelligent retrieval
+- Contextual relationships that help AI understand relevance
+- Pattern recognition data that enables proactive suggestions
+
+**Human-Extractable Insights**:
+- While knowledge is stored for AI consumption, humans can extract valuable insights through:
+  - **Web Dashboard**: Browse and search through accumulated knowledge
+  - **Pattern Recognition**: Identify recurring themes and solutions across projects
+  - **Knowledge Synthesis**: Derive high-level principles from detailed implementation knowledge
+  - **Learning Acceleration**: Understand team knowledge and best practices through accumulated insights
+
+### 🎯 **Value-Driven Knowledge Capture**
+
+**Maximizing Knowledge ROI**: By focusing on usage frequency and value density:
+
+- **Contextual Triggering**: Knowledge is captured and surfaced based on actual development context
+- **Relevance Scoring**: Only valuable, applicable knowledge reaches the developer
+- **Automatic Organization**: AI-driven classification ensures knowledge remains findable and useful
+- **Continuous Refinement**: Knowledge quality improves through usage patterns and feedback loops
+
+This philosophy transforms knowledge from a documentation burden into an intelligent development accelerator that grows more valuable with every coding session.
+
 ## Installation
 
 ### Prerequisites
@@ -96,6 +147,39 @@ The web interface provides comprehensive access to your knowledge database:
 
 The web interface transforms your local knowledge database into an accessible, browsable resource that helps you leverage your accumulated development experience effectively.
 
+## How it Works
+
+Knowledge Bank integrates seamlessly with Claude Code to automatically capture, organize, and retrieve development knowledge through a simple three-step process:
+
+### 📥 **Knowledge Capture**
+- **Automatic Collection**: Uses Claude Code's hook system to capture development insights during your coding sessions
+- **Smart Classification**: Organizes knowledge by type (code patterns, architecture, configurations, pitfalls, etc.)
+- **Context Awareness**: Associates knowledge with specific projects, files, and development activities
+
+### 🗄️ **Knowledge Storage**
+- **Local Database**: Stores all knowledge in a local SQLite database for privacy and speed
+- **Rich Organization**: Categorizes knowledge by scope, source, and development lifecycle stage
+- **Full-Text Search**: Enables fast searching across all stored knowledge items
+
+### 🔍 **Knowledge Retrieval**
+- **Intelligent Search**: Finds relevant knowledge based on your current development context
+- **Multiple Interfaces**: Access through Claude Code (via MCP), web dashboard, or command line
+- **Context Injection**: Automatically surfaces relevant insights when you need them
+
+### 🔧 **Integration Points**
+
+**Claude Code Integration**
+- Hooks into Claude Code's session lifecycle for automatic knowledge capture
+- Provides MCP server for real-time knowledge search and retrieval
+- Works transparently without disrupting your development workflow
+
+**Web Dashboard**
+- Browse and explore your knowledge database through a user-friendly interface
+- Advanced search and filtering capabilities
+- Session history and hook event tracking
+
+The system transforms your development sessions into a continuously growing, searchable knowledge base that enhances productivity and preserves valuable insights for future reference.
+
 ## License
 
 This project is licensed under the ISC License - see the [LICENSE](LICENSE) file for details.

package/dist/claude-marketplace/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
       "name": "core",
       "source": "./core",
       "description": "Core plugins for knowledge bank",
-      "version": "0.1.2-alpha.276",
+      "version": "0.1.2-alpha.312",
       "author": {
         "name": "garbywu",
         "email": "gabrywu@apache.org"
@@ -18,6 +18,6 @@
   ],
   "metadata": {
     "description": "A marketplace for knowledge bank plugins by gabrywu",
-    "version": "0.1.2-alpha.276"
+    "version": "0.1.2-alpha.312"
   }
 }

package/dist/claude-marketplace/core/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "core",
   "description": "Core plugins for knowledge bank",
-  "version": "0.1.2-alpha.276",
+  "version": "0.1.2-alpha.312",
   "author": {
     "name": "garbywu",
     "email": "gabrywu@apache.org"

package/dist/claude-marketplace/core/hooks/hooks.json

@@ -102,6 +102,78 @@
           }
         ]
       }
+    ],
+    "SubagentStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank subagent-start"
+          }
+        ]
+      }
+    ],
+    "PostToolUseFailure": [
+      {
+        "matcher": "*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank post-tool-use-failure"
+          }
+        ]
+      }
+    ],
+    "TeammateIdle": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank teammate-idle"
+          }
+        ]
+      }
+    ],
+    "TaskCompleted": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank task-completed"
+          }
+        ]
+      }
+    ],
+    "ConfigChange": [
+      {
+        "matcher": "*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank config-change"
+          }
+        ]
+      }
+    ],
+    "WorktreeCreate": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank worktree-create"
+          }
+        ]
+      }
+    ],
+    "WorktreeRemove": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx knowledge-bank worktree-remove"
+          }
+        ]
+      }
     ]
   }
 }

package/dist/claude-marketplace/core/skills/knowledge.add/SKILL.md

@@ -1,15 +1,22 @@
 ---
 name: knowledge.add
-description: Add a new knowledge item to the knowledge base with automatic content generation from user instruction
-disable-model-invocation: false
-user-invocable: true
-allowed-tools: knowledge_create, knowledge_search, knowledge_list, Write, Read, Bash, Skill
+description: Create new knowledge items with intelligent content generation from user instructions. Use when users say 'document this', 'add to knowledge base', 'remember this for later', 'save this insight', 'capture this solution', or 'make a note of this'. Also trigger when users complete complex troubleshooting, discover useful patterns, make architectural decisions, find implementation insights worth capturing, or encounter solutions to tricky problems, even if they don't explicitly mention 'knowledge' or 'documentation'. Automatically validates quality, generates comprehensive content, and includes relevant code examples and file references.
+disable-model-invocation: false # Allows the skill to call Claude for content generation
+user-invocable: true # Makes this skill available to users via /knowledge-add
+allowed-tools: knowledge_create, knowledge_search, knowledge_list, Write, Read, Bash, Skill # MCP tools and system access for knowledge operations
 ---
 
 # Knowledge Addition
 
 Add a new knowledge item to the knowledge base with automatic content generation from user instruction. This skill uses the `knowledge-bank-management` MCP server to interact with the knowledge base.
 
+**System Environment Variables:**
+- CLAUDE_SESSION_ID=!`echo $CLAUDE_SESSION_ID`
+- CLAUDE_SESSION_SOURCE=!`echo $CLAUDE_SESSION_SOURCE`
+- CLAUDE_SESSION_MODE=!`echo $CLAUDE_SESSION_MODE`
+- CLAUDE_SESSION_REPOSITORYID=!`echo $CLAUDE_SESSION_REPOSITORYID`
+- CLAUDE_SESSION_CWD=!`echo $CLAUDE_SESSION_CWD`
+
 ## Knowledge Detection Integration
 
 **IMPORTANT:** Use the `knowledge.detect` skill to analyze whether the user instruction contains detectable knowledge points before proceeding with knowledge creation. This ensures alignment with established knowledge standards and prevents creation of low-value entries.
@@ -39,7 +46,20 @@ To use knowledge detection:
 
 Use the knowledge detection results from `knowledge.detect` to guide content generation, ensuring adherence to established knowledge categories, quality standards, and structure requirements.
 
-## Usage Example
+## Examples
+
+**Example 1:**
+Input: "Document how we handle error logging in this project"
+Output: Creates a `code_pattern` knowledge item titled "Error Logging Pattern with Winston" including:
+- Analysis of `src/utils/logger.ts` implementation
+- Code examples for error handling patterns
+- File references to relevant logging modules
+- Tagged with 'logging', 'winston', 'error-handling'
 
-- User: `/knowledge-add "Document how we handle error logging in this project"`
-- Result: Analyzes logging implementation, scans `logger.ts`, and creates a new `code_pattern` knowledge item.
+**Example 2:**
+Input: "Remember this authentication middleware approach we just implemented"
+Output: Creates an `architecture` knowledge item titled "JWT Authentication Middleware Pattern" including:
+- Explanation of middleware chain implementation
+- Code examples from `src/middleware/auth.ts`
+- Security considerations and best practices
+- Integration points with user sessions

package/dist/claude-marketplace/core/skills/knowledge.collect/SKILL.md

@@ -1,11 +1,11 @@
 ---
 name: knowledge.collect
-description: Intelligent knowledge management with CRUD operations and lifecycle management
-disable-model-invocation: false
-user-invocable: true
-allowed-tools: knowledge_create, knowledge_get, knowledge_update, knowledge_delete, knowledge_list, knowledge_search, knowledge_update_status, knowledge_stats, Write, Read, Bash, Skill
-context: fork
-agent: general-purpose
+description: Automatically capture and process knowledge from completed development sessions with intelligent deduplication and quality control. Use proactively after task completion when users finish debugging, complete features, solve complex problems, or wrap up development sessions. Also trigger when users say 'save this session', 'capture what we learned', 'collect insights', or when significant technical work concludes with valuable discoveries, successful problem resolution, or architectural decisions that should be preserved for future reference.
+disable-model-invocation: false # Enables automated knowledge detection and processing
+user-invocable: true # Available for explicit collection requests
+allowed-tools: knowledge_create, knowledge_get, knowledge_update, knowledge_delete, knowledge_list, knowledge_search, knowledge_update_status, knowledge_stats, Write, Read, Bash, Skill # Full MCP suite for comprehensive knowledge management
+context: fork # Executes in isolated context to avoid interference with main session
+agent: general-purpose # Uses general-purpose agent for broad analytical capabilities
 ---
 
 # Knowledge Collection & Management
@@ -13,11 +13,12 @@ agent: general-purpose
 Intelligent knowledge lifecycle management system that processes detected knowledge items through create/update/delete operations with smart deduplication and quality control.
 
 **Important:** You have access to the `knowledge-bank-management` MCP server which provides all necessary tools for knowledge management operations. Use these MCP tools to interact with the knowledge bank instead of CLI commands.
-**System Environment Variables:** 
+**System Environment Variables:**
 - CLAUDE_SESSION_ID=!`echo $CLAUDE_SESSION_ID`
 - CLAUDE_SESSION_SOURCE=!`echo $CLAUDE_SESSION_SOURCE`
 - CLAUDE_SESSION_MODE=!`echo $CLAUDE_SESSION_MODE`
 - CLAUDE_SESSION_REPOSITORYID=!`echo $CLAUDE_SESSION_REPOSITORYID`
+- CLAUDE_SESSION_CWD=!`echo $CLAUDE_SESSION_CWD`
 
 ## File Parameter Configuration
 
@@ -54,27 +55,63 @@ Use the `knowledge.detect` skill results to identify valuable knowledge items ba
 
 When valuable knowledge is detected, output it in this structured format:
 
-### Output Template
+### Output Format
 
+ALWAYS use this exact template:
+
+When Knowledge is Collected:
+```
+🧠 KNOWLEDGE COLLECTED
+
+**Total Processed:** [X] items
+**Created:** [Y] new items
+**Updated:** [Z] existing items
+
+1. **[Knowledge Title]**
+   Type: [knowledge_type]
+   Action: [Created|Updated|Merged]
+   Files: [Relevant file paths]
+
+[Additional collected items...]
+```
+
+When No Valuable Knowledge is Found:
+```
+📝 NO SIGNIFICANT KNOWLEDGE DETECTED
+
+**Reason:** [Brief explanation - e.g., "Routine operations without project-specific insights"]
 ```
-🧠 KNOWLEDGE DETECTED
 
-Title: [Clear, descriptive title]
-Type: [architecture|code_pattern|pitfall|tool_usage|api_usage|exploration]
-Summary: [One-sentence description of the insight]
+## Examples
 
-Files: [Comma-separated list of relevant file paths]
-Tags: [Searchable keywords for future reference]
+**Example 1:**
+Input: "Collect knowledge from our React debugging session"
+Output:
 ```
+🧠 KNOWLEDGE COLLECTED
 
-### No Knowledge Found Output
+**Total Processed:** 3 items
+**Created:** 2 new items
+**Updated:** 1 existing items
 
-When no valuable knowledge is detected:
+1. **React useState Re-rendering Debug Pattern**
+   Type: pitfall
+   Action: Created
+   Files: src/components/UserProfile.tsx, src/hooks/useUserData.ts
 
+2. **useMemo Performance Optimization**
+   Type: code_pattern
+   Action: Updated
+   Files: src/components/UserProfile.tsx
+```
+
+**Example 2:**
+Input: "Process the configuration file updates we just made"
+Output:
 ```
 📝 NO SIGNIFICANT KNOWLEDGE DETECTED
 
-Reason: [Brief explanation - e.g., "Routine operations without project-specific insights"]
+**Reason:** Standard configuration updates without novel patterns or complex problem-solving insights
 ```
 
 ## 2. Duplicate Detection & Conflict Resolution
@@ -85,9 +122,13 @@ Reason: [Brief explanation - e.g., "Routine operations without project-specific
 
 #### Step 1: Search for Essential Fields
 Use `knowledge_search` to get potential duplicates:
+**IMPORTANT**: The `knowledge_search` tool now requires a `cwd` parameter. The `cwd` value must be taken from !`echo $CLAUDE_SESSION_CWD`:
+
+
 ```
 knowledge_search({
   query: ["extracted", "keywords", "from", "new", "knowledge"],
+  cwd: "!`echo $CLAUDE_SESSION_CWD`",
   to: "duplicate-search.json"
 })
 ```
@@ -157,7 +198,7 @@ To ensure reliability and handle large content, always use file-based operations
 # Workflow Examples
 
 ## Create New Knowledge
-1. Use `knowledge_search` with the `to` parameter (filename only, e.g., `search-results.json`) to check for duplicates and save results to the !`npx knowledge-bank config -t` directory.
+1. Use `knowledge_search` with the `cwd: !`echo $CLAUDE_SESSION_CWD`` and `to` parameter (filename only, e.g., `search-results.json`) to check for duplicates and save results to the !`npx knowledge-bank config -t` directory.
 2. If no duplicates found:
    - Write new knowledge data to a local JSON file in !`npx knowledge-bank config -t`.
    - Use `knowledge_create` with the `from` parameter (filename only, e.g., `new-knowledge.json`) and `session_id`,which is ${CLAUDE_SESSION_ID}, and `repository_id`, which is !`echo $CLAUDE_SESSION_REPOSITORYID`.
@@ -177,7 +218,7 @@ To ensure reliability and handle large content, always use file-based operations
 3. Use `knowledge_update_status` to promote to verified
 
 ## Search and Retrieve
-1. Use `knowledge_search` with relevant terms
+1. Use `knowledge_search` with `cwd: !`echo $CLAUDE_SESSION_CWD`` and relevant terms
 2. Use `knowledge_get` to retrieve full details
 3. Use returned knowledge to enhance current context
 

package/dist/claude-marketplace/core/skills/knowledge.delete/SKILL.md

@@ -1,22 +1,68 @@
 ---
 name: knowledge.delete
-description: Delete a specific knowledge item from the knowledge base
-disable-model-invocation: false
-user-invocable: true
-allowed-tools: knowledge_delete, knowledge_get, knowledge_search, Write, Read, Bash
-context: fork
-agent: general-purpose
+description: Safely remove specific knowledge items from the knowledge base with verification and confirmation. Use when users say 'delete knowledge item', 'remove this knowledge', 'clean up outdated entries', or specify knowledge IDs for removal. Also trigger for maintenance operations when cleaning up duplicate entries after merges, removing obsolete information that no longer applies, deleting low-quality or incorrectly categorized knowledge items, or when users explicitly request removal of specific knowledge entries by ID or description.
+disable-model-invocation: false # Enables content analysis for verification before deletion
+user-invocable: true # Available for explicit deletion requests
+allowed-tools: knowledge_delete, knowledge_get, knowledge_search, Write, Read, Bash # MCP tools for safe deletion operations with verification
+context: fork # Isolated execution for safe deletion operations
+agent: general-purpose # Handles verification and confirmation workflows
 ---
 
 # Knowledge Deletion
 
 Remove a specific knowledge item from the knowledge base using its ID.
 
+**System Environment Variables:**
+- CLAUDE_SESSION_ID=!`echo $CLAUDE_SESSION_ID`
+- CLAUDE_SESSION_SOURCE=!`echo $CLAUDE_SESSION_SOURCE`
+- CLAUDE_SESSION_MODE=!`echo $CLAUDE_SESSION_MODE`
+- CLAUDE_SESSION_REPOSITORYID=!`echo $CLAUDE_SESSION_REPOSITORYID`
+- CLAUDE_SESSION_CWD=!`echo $CLAUDE_SESSION_CWD`
+
 ## Workflow
 
-1. **Identification**: Identify the knowledge item to delete using the ID provided in `$ARGUMENTS`. If the ID is not clear, use `knowledge_search` to find it.
+1. **Identification**: Identify the knowledge item to delete using the ID provided in `$ARGUMENTS`. If the ID is not clear, use `knowledge_search` with cwd: !`echo $CLAUDE_SESSION_CWD` to find it.
 2. **Verification**: Use the `knowledge_get` tool to retrieve the details of the item to ensure it exists and to show the user what will be deleted.
 3. **Execution**: Use the `knowledge_delete` MCP tool with the specified ID.
 4. **Reporting**: Confirm successful deletion or report any errors.
 
 ⚠️ **Warning**: Deletion is permanent and cannot be undone.
+
+## Output Format
+
+ALWAYS use this exact template:
+
+```
+🗑️ KNOWLEDGE DELETION COMPLETED
+
+**Deleted Item:**
+- ID: [X]
+- Title: [Title]
+- Type: [knowledge_type]
+- Status: [status]
+
+**Confirmation:** Knowledge item successfully removed from database
+**Impact:** [Brief description of what was removed]
+```
+
+## Examples
+
+**Example 1:**
+Input: "Delete knowledge item 42"
+Output:
+```
+🗑️ KNOWLEDGE DELETION COMPLETED
+
+**Deleted Item:**
+- ID: 42
+- Title: Outdated React Class Component Pattern
+- Type: code_pattern
+- Status: draft
+
+**Confirmation:** Knowledge item successfully removed from database
+**Impact:** Removed deprecated React pattern that's no longer recommended
+```
+
+**Example 2:**
+Input: "Remove duplicate authentication knowledge"
+Output: [First searches for and identifies the duplicate, then shows deletion confirmation]

package/dist/claude-marketplace/core/skills/knowledge.detect/SKILL.md

@@ -1,9 +1,9 @@
 ---
 name: knowledge.detect
-description: Define and collect technical knowledge points from the current conversation
-disable-model-invocation: false
-user-invocable: true
-allowed-tools: Read, Write, Bash
+description: Analyze current conversation session to identify and validate high-value technical knowledge points using established criteria. Use when users ask 'what knowledge can we capture here?', 'is this worth documenting?', 'analyze this session for insights', or when you need to assess knowledge quality before manual capture, during content validation workflows, or when determining if technical discussions contain documentable insights. Also trigger when reviewing completed development sessions, evaluating problem-solving conversations, or validating content before knowledge creation operations.
+disable-model-invocation: false # Enables AI analysis of conversation content for knowledge extraction
+user-invocable: true # Available to users for manual knowledge quality assessment
+allowed-tools: Read, Write, Bash # File system access for session analysis and temporary data processing
 ---
 
 # In-Session Knowledge Detection & Collection
@@ -90,26 +90,58 @@ For each qualified insight, extract:
 
 ## 8. Collection Results Output
 
-### When Knowledge Points Are Detected
+### Output Format
 
+ALWAYS use this exact template:
+
+When Knowledge Points Are Detected:
 ```
 🧠 Knowledge Detected
 
 **Total:** [Number] point(s)
 
 1. **[Knowledge Title]**
-   - Category: [category] | Scope: [scope]
-   - Location: [relevant files/components]
-   - Elements:
+   Type: [knowledge_type]
+   Summary: [One-sentence description]
+   Files: [Relevant file paths]
+   Elements:
      - **What**: [insight]
      - **Why**: [rationale]
      - **How**: [implementation]
 ```
 
-### When No Knowledge Points Are Detected
-
+When No Knowledge Points Are Detected:
 ```
 ❌ No Knowledge Detected
 
 **Reason:** [Brief explanation based on qualification criteria]
 ```
+
+## Examples
+
+**Example 1:**
+Input: "Analyze our conversation about debugging the React state management issue"
+Output:
+```
+🧠 Knowledge Detected
+
+**Total:** 1 point(s)
+
+1. **React State Management Debugging Pattern**
+   Type: pitfall
+   Summary: Step-by-step approach to diagnose useState re-rendering issues
+   Files: src/components/UserProfile.tsx, src/hooks/useUserData.ts
+   Elements:
+     - **What**: Identified unnecessary re-renders caused by object recreation in useState
+     - **Why**: Component re-rendered on every parent update due to new object reference
+     - **How**: Use useMemo for object values and useCallback for event handlers
+```
+
+**Example 2:**
+Input: "Check if our discussion about updating package.json has any knowledge value"
+Output:
+```
+❌ No Knowledge Detected
+
+**Reason:** Routine package management without project-specific insights or novel problem-solving approaches
+```

package/dist/claude-marketplace/core/skills/knowledge.enhance/SKILL.md

@@ -1,9 +1,9 @@
 ---
 name: knowledge.enhance
-description: Search and inject relevant knowledge into current session based on user input
-disable-model-invocation: false
-user-invocable: true
-allowed-tools: knowledge_search, knowledge_list, knowledge_get, Read, Write, Bash
+description: Search and inject relevant knowledge from past sessions when users ask development questions, encounter errors, need architecture guidance, or want best practices. Use this proactively whenever users mention 'how do I', 'I'm having trouble with', 'error', 'bug', 'stuck', 'implementation challenge', 'architecture decision', 'best practice', or describe specific technical problems. Also trigger when users reference files, frameworks, or technologies that may have existing knowledge, ask about patterns or solutions, or need contextual help with coding tasks, even if they don't explicitly request knowledge lookup.
+disable-model-invocation: false # Enables intelligent keyword extraction and relevance analysis
+user-invocable: true # Available for explicit knowledge enhancement requests
+allowed-tools: knowledge_search, knowledge_list, knowledge_get, Read, Write, Bash # MCP tools for knowledge retrieval and file operations
 ---
 
 # Knowledge Enhancement & Context Injection
@@ -16,6 +16,7 @@ Intelligent knowledge retrieval system that summarizes relevant keywords based o
 - CLAUDE_SESSION_SOURCE=!`echo $CLAUDE_SESSION_SOURCE`
 - CLAUDE_SESSION_MODE=!`echo $CLAUDE_SESSION_MODE`
 - CLAUDE_SESSION_REPOSITORYID=!`echo $CLAUDE_SESSION_REPOSITORYID`
+- CLAUDE_SESSION_CWD=!`echo $CLAUDE_SESSION_CWD`
 
 ## File Parameter Configuration
 
@@ -61,9 +62,12 @@ Use `knowledge_list` for type-specific knowledge retrieval with filters.
 #### Step 1: Search for Essential Fields
 Use `knowledge_search` or `knowledge_list` to get essential fields only:
 
+**IMPORTANT**: The `knowledge_search` tool now requires a `cwd` parameter. The `cwd` value must be taken from !`echo $CLAUDE_SESSION_CWD`:
+
 ```
 knowledge_search({
   query: ["extracted", "keywords"],
+  cwd: "!`echo $CLAUDE_SESSION_CWD`",
   to: "search-results.json"
 })
 ```
@@ -96,13 +100,14 @@ knowledge_get({
 ```
 
 #### Step 4: Context Injection Format
-Present retrieved knowledge in this format:
+
+ALWAYS use this exact template:
 
 ```
 🔍 RELEVANT KNOWLEDGE FOUND
 
-**Relevance Analysis:** Found X relevant items from Y search results
-**Top Relevance Scores:** [item1: 0.85, item2: 0.72, item3: 0.61]
+**Relevance Analysis:** Found [X] relevant items from [Y] search results
+**Top Relevance Scores:** [item1: 0.XX, item2: 0.XX, item3: 0.XX]
 
 <knowledge_content>
 [Insert the FULL content of the retrieved knowledge files here]
@@ -117,6 +122,29 @@ Present retrieved knowledge in this format:
 ### No Results Handling
 If no relevant knowledge found, respond with empty string, indicating no context to inject.
 
+## Examples
+
+**Example 1:**
+Input: "I'm having trouble with React useState causing infinite re-renders"
+Output:
+```
+🔍 RELEVANT KNOWLEDGE FOUND
+
+**Relevance Analysis:** Found 2 relevant items from 15 search results
+**Top Relevance Scores:** [React State Debugging: 0.89, useState Hook Patterns: 0.74]
+
+<knowledge_content>
+# React State Management Debugging Pattern
+...
+[Full knowledge content with debugging steps and solutions]
+...
+</knowledge_content>
+```
+
+**Example 2:**
+Input: "How do I configure webpack for this project?"
+Output: [Empty response if no webpack-related knowledge exists]
+
 ## 4. Enhancement Strategies
 
 ### Proactive Knowledge Discovery