add tools for self improvement

CLAUDE.md

@@ -41,3 +41,95 @@ The project roadmap is maintained in `ROADMAP.md`. When working on features or d
 - Update the roadmap when completing items (move to "Completed" section)
 - Add new items as they are identified or requested
 - Keep items well-organized by category (Performance, Features, etc.)
+
+## Elroy Agent Tools (Also Available as Claude Code Skills)
+
+Elroy provides tools that are available both to the Elroy agent directly AND as Claude Code skills (via `/` commands).
+
+### Memory Management Tools
+
+These tools allow creating and managing long-term memories:
+
+- **create_memory(name, text)** - Create a new long-term memory
+  - Claude Code: `/remember "content"`
+  - Example: `create_memory("Project preferences", "User prefers TypeScript")`
+
+- **examine_memories(query)** - Search through memories
+  - Claude Code: `/recall "query"`
+  - Example: `examine_memories("authentication method")`
+
+- **print_memories()** - List all active memories
+  - Claude Code: `/list-memories`
+
+- **create_reminder(name, text, trigger_datetime, reminder_context)** - Create a reminder
+  - Claude Code: `/remind "content"`
+  - Example: `create_reminder("Review PR", "Check the authentication PR", trigger_datetime="2025-01-26 14:00")`
+
+- **print_active_reminders()** - List active reminders
+  - Claude Code: `/list-reminders`
+
+- **ingest_doc(path)** - Ingest documents into memory
+  - Claude Code: `/ingest path`
+  - Example: `ingest_doc("docs/architecture.md")`
+
+### Self-Improvement Tools
+
+These tools enable Elroy to understand and improve itself:
+
+- **introspect(query)** - Ask questions about Elroy's implementation
+  - Claude Code: `/introspect "query"`
+  - Example: `introspect("How does memory consolidation work?")`
+  - Returns: Detailed explanation with file paths and implementation details
+
+- **make_improvement(description, create_branch, run_tests, submit_pr)** - Implement an improvement
+  - Claude Code: `/make-improvement "description"`
+  - Example: `make_improvement("Add date-aware search to examine_memories")`
+  - Workflow: Plan → Implement → Test → Document → Submit PR
+  - Returns: PR URL when submitted
+
+- **review_roadmap()** - Review roadmap, issues, and recent commits
+  - Example: `review_roadmap()`
+  - Returns: Current project status and priorities
+
+### Usage in Elroy Agent
+
+When working as the Elroy agent (not in Claude Code), you can call these tools directly:
+
+```python
+# Understand implementation before making changes
+result = introspect("How does the tool registry work?")
+
+# Check project priorities
+status = review_roadmap()
+
+# Implement an improvement
+pr_url = make_improvement(
+    "Add better error messages for tool failures",
+    create_branch=True,
+    run_tests=True,
+    submit_pr=True
+)
+```
+
+### Usage in Claude Code
+
+When working in Claude Code, use the slash command versions:
+
+```bash
+/introspect "How does memory consolidation work?"
+/make-improvement "Add date-aware search"
+/recall "project architecture decisions"
+/remember "User prefers functional components"
+```
+
+### Development Workflow
+
+The self-improvement tools create a complete feedback loop:
+
+1. **Identify improvement** - Through usage patterns, user request, or review_roadmap()
+2. **Understand current state** - Use introspect() to explore implementation
+3. **Plan changes** - Review ROADMAP.md and issues with review_roadmap()
+4. **Implement** - Use make_improvement() for structured implementation
+5. **Submit for review** - PR created automatically with tests and documentation
+
+This enables Elroy to actively participate in its own development.

README.md

@@ -134,6 +134,8 @@ Elroy provides skills for [Claude Code](https://github.com/anthropics/claude-cod
 - `/remind` - Create a reminder
 - `/list-reminders` - List active reminders
 - `/ingest` - Ingest documents into memory
+- `/introspect` - Ask questions about Elroy's implementation (for development)
+- `/make-improvement` - Implement a feature or improvement and submit a PR (for development)
 
 ### Installation
 

ROADMAP.md

@@ -8,7 +8,16 @@ This document tracks planned improvements and features for Elroy.
 
 ## Future Items
 
-(Items will be added here as they are identified)
+### Self-Improvement Framework (Partial)
+Enable Elroy to actively participate in its own development:
+- **Usage analytics tracking**: Capture interaction outcomes (success/failure/retry patterns)
+- **Analysis tools**: Tools to query usage patterns, review roadmap
+- **Improvement proposals**: System for agent to suggest and prioritize improvements based on observed data
+- **Development session mode**: Special mode where agent works on implementing improvements
+
+**Completed foundation**:
+- ✅ `/introspect` skill - Agent can understand its own implementation
+- ✅ `/make-improvement` skill - Agent can implement changes and submit PRs
 
 ## Completed
 
@@ -22,12 +31,16 @@ This document tracks planned improvements and features for Elroy.
   - Configured automatic logging for slow operations (>100ms)
 
 ### Developer Experience
-- **Create Claude Code skills for memory tools** (Completed: 2025-01)
-  - Built complete set of 6 Claude Code skills in `claude-skills/` directory
-  - `/remember` - Create long-term memories
-  - `/recall` - Search through memories
-  - `/list-memories` - List all active memories
-  - `/remind` - Create reminders
-  - `/list-reminders` - List active reminders
-  - `/ingest` - Ingest documents into memory
-  - Includes installation script with help documentation
+- **Create Claude Code skills for memory and development tools** (Completed: 2025-01)
+  - Built complete set of 8 Claude Code skills in `claude-skills/` directory
+  - Memory management:
+    - `/remember` - Create long-term memories
+    - `/recall` - Search through memories
+    - `/list-memories` - List all active memories
+    - `/remind` - Create reminders
+    - `/list-reminders` - List active reminders
+    - `/ingest` - Ingest documents into memory
+  - Self-improvement capabilities:
+    - `/introspect` - Ask questions about Elroy's implementation (enables self-awareness)
+    - `/make-improvement` - Implement features/improvements and submit PRs (enables self-modification)
+  - Includes installation script and comprehensive documentation

claude-skills/README.md

@@ -6,6 +6,8 @@ This directory contains Claude Code skills that expose Elroy's memory management
 
 Elroy skills allow you to use Elroy's powerful memory management features directly from within Claude Code sessions. This creates a seamless integration between Claude Code and Elroy's long-term memory capabilities.
 
+Additionally, development skills (`/introspect` and `/make-improvement`) enable Elroy to understand and improve itself - creating a self-improvement feedback loop.
+
 ## Available Skills
 
 | Skill | Description | Usage |
@@ -16,6 +18,8 @@ Elroy skills allow you to use Elroy's powerful memory management features direct
 | `/remind` | Create a reminder | `/remind "Review PR tomorrow"` |
 | `/list-reminders` | List active reminders | `/list-reminders` |
 | `/ingest` | Ingest documents | `/ingest docs/` |
+| `/introspect` | Ask about Elroy's implementation | `/introspect "How does memory consolidation work?"` |
+| `/make-improvement` | Implement and PR a feature | `/make-improvement "Add date-aware search"` |
 
 ## Installation
 
@@ -111,6 +115,26 @@ Make documentation available to Elroy for context:
 /ingest specs/api-documentation.pdf
 ```
 
+### Self-Improvement Workflow
+
+Use the development skills to understand and improve Elroy:
+
+```bash
+# Understand how something works
+/introspect "How does memory consolidation work?"
+
+# Implement an improvement
+/make-improvement "Add date-aware search to examine_memories"
+
+# The agent will:
+# 1. Use /introspect to understand current implementation
+# 2. Plan the change and ask for approval
+# 3. Implement, test, and document the change
+# 4. Submit a pull request
+```
+
+This creates a complete feedback loop where Elroy can understand and improve itself.
+
 ## How It Works
 
 Each skill is a Claude Code skill (SKILL.md file with YAML frontmatter) that:

claude-skills/install-skills.sh

@@ -35,12 +35,14 @@ OPTIONS:
     --help, -h        Show this help message
 
 INSTALLED SKILLS:
-    /remember       - Create a long-term memory
-    /recall         - Search through memories
-    /list-memories  - List all memories
-    /remind         - Create a reminder
-    /list-reminders - List active reminders
-    /ingest         - Ingest documents into memory
+    /remember        - Create a long-term memory
+    /recall          - Search through memories
+    /list-memories   - List all memories
+    /remind          - Create a reminder
+    /list-reminders  - List active reminders
+    /ingest          - Ingest documents into memory
+    /introspect      - Ask questions about Elroy's implementation
+    /make-improvement - Implement a feature or improvement and submit a PR
 
 REQUIREMENTS:
     - Elroy must be installed and available in PATH
@@ -101,6 +103,8 @@ SKILLS=(
     "remind"
     "list-reminders"
     "ingest"
+    "introspect"
+    "make-improvement"
 )
 
 # Uninstall if requested

claude-skills/introspect/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: introspect
+description: Ask questions about Elroy's implementation and codebase
+disable-model-invocation: false
+---
+
+Ask questions about how Elroy is implemented. This skill helps you understand Elroy's codebase, architecture, and implementation details.
+
+## Usage
+
+When the user invokes this skill with `/introspect [QUERY]`, you should use the Task tool with the Explore agent to search through the Elroy codebase.
+
+**IMPORTANT**: You are currently working in the Elroy repository at `/Users/tombedor/development/elroy`. Use the Task tool to explore the codebase.
+
+### Example Task Tool Usage
+
+```
+Task tool with:
+- subagent_type: "Explore"
+- description: "Explore Elroy codebase"
+- prompt: "<the user's query about implementation>"
+- model: "sonnet" (for speed)
+```
+
+### What You Can Ask About
+
+- **Architecture**: "How does memory consolidation work?"
+- **Implementation**: "Where is the memory recall classifier implemented?"
+- **Data models**: "What database tables exist?"
+- **Tools**: "How are tools registered and executed?"
+- **Configuration**: "What configuration options are available?"
+- **Features**: "How do reminders work?"
+- **Performance**: "What latency tracking exists?"
+
+### Examples
+
+User query: `/introspect "How does memory consolidation work?"`
+
+You should:
+1. Use the Task tool with Explore agent to search the codebase
+2. Look for files related to consolidation (e.g., consolidation.py)
+3. Examine relevant code
+4. Provide a clear explanation of how it works
+
+User query: `/introspect "Where is the latency tracker used?"`
+
+You should:
+1. Use Task/Explore to find latency.py and usage sites
+2. Show where it's initialized and tracked
+3. Explain the tracking mechanism
+
+### Response Format
+
+Provide:
+1. **Summary**: Brief answer to the question
+2. **Key Files**: List relevant file paths with line numbers
+3. **How It Works**: Explanation of the implementation
+4. **Related Code**: Any connected systems or patterns
+
+### Repository Structure
+
+The Elroy codebase is organized as:
+- `elroy/core/` - Core infrastructure (logging, latency, context, tracing)
+- `elroy/db/` - Database models and operations
+- `elroy/repository/` - Data access layer (memories, reminders, messages)
+- `elroy/tools/` - Agent tools and commands
+- `elroy/messenger/` - Message processing and agent loop
+- `elroy/cli/` - CLI interface
+- `elroy/config/` - Configuration management
+- `elroy/io/` - Input/output handling
+- `claude-skills/` - Claude Code integration skills
+
+### Notes
+
+- Use the Explore agent for thorough searches across the codebase
+- Include file paths with line numbers (e.g., `elroy/core/latency.py:45`)
+- Explain not just "what" but "why" when possible
+- Reference related systems and how they connect
+
+This skill enables self-awareness - helping the agent understand its own implementation.