refactor: Extract context management to reusable tools + Simplify profile API (#1587)

* refactor: Simplify profile API - Remove URI scheme for built-in profiles

Fixes #1575

## Changes

Simplified the profile API so built-in profiles can be referenced by simple
names instead of requiring URI schemes. Reduces unnecessary complexity while
maintaining full backward compatibility.

### API Simplification

**Before** (Complex):
```python
loader.load("amplihack://profiles/coding")  # Built-in
loader.load("file:///path/to/custom.yaml")  # Custom
```

**After** (Simple):
```python
loader.load("coding")                      # Built-in - just the name!
loader.load("file:///path/to/custom.yaml") # Custom - unchanged
loader.load("amplihack://profiles/coding") # Still works (backward compat)
```

### Implementation Details

**loader.py** (~5 lines):
- Added simple scheme detection: if no "://" present, treat as built-in name
- Reuses existing _load_builtin() method for both paths
- Maintains backward compatibility for amplihack:// URIs

**config.py** (~4 lines):
- Changed default from "amplihack://profiles/all" to "all"
- Updated docstrings and examples

**cli.py** (~8 lines):
- Changed list_profiles() to display simple names instead of amplihack:// URIs
- Added proper try/except for optional rich dependency
- More user-friendly output

**README.md** (~20 lines):
- Added section showing both syntaxes with clear preference for simple names
- Documented backward compatibility
- Updated all examples to prefer simple syntax

**test_simple_names.py** (NEW):
- 6 comprehensive test cases for simple name API
- Tests backward compatibility
- Tests edge cases

### Testing

✅ Simple names work: load("coding"), load("all"), load("research")
✅ Backward compat: load("amplihack://profiles/coding") still works
✅ File scheme unchanged: load("file:///path") unchanged
✅ Edge cases: load("all.yaml") works correctly
✅ All manual tests passed

### Philosophy Alignment

**Ruthless Simplicity**: Removes unnecessary amplihack:// scheme complexity
through a 5-line string check, while maintaining backward compatibility.

**Modular Design**: Changes localized to logical boundaries, no coupling.

**Zero-BS**: All code works, no stubs or placeholders.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* refactor: Extract context management to reusable tools following best practices

Refactored context management system to follow Claude Code skill best practices
and the brick philosophy. Separated business logic from presentation layer into
reusable, self-contained tools.

Key Changes:
- Created context_manager.py (900+ lines) - Consolidates token monitoring, context
  extraction, snapshot creation, and rehydration
- Created transcript_manager.py (577 lines) - Consolidates transcript operations
- Created tool_registry.py (333 lines) - Extensible hook registration system
- Created context_automation_hook.py (190 lines) - Bridge between hooks and tools
- Refactored post_tool_use.py - Uses registry pattern instead of inline logic
- Reduced SKILL.md from 513→395 lines, instructions only (< 500 ✓)
- Converted transcripts.py to transcripts.md (markdown instructions)

Benefits:
- Clean separation of concerns (tools/ vs skills/ vs commands/)
- Extensible hook system (easy to add new tool hooks)
- Follows Claude Code best practices (< 500 line skills, instructions-only)
- Self-contained, regeneratable modules (brick philosophy)
- Zero-BS implementation (all functions work completely)

Architecture:
- Business logic in .claude/tools/amplihack/ (reusable bricks)
- Skills/commands are instruction-only markdown
- Hook system supports multiple tools via registry pattern
- See .claude/docs/context-management-refactoring-architecture.md

Code Review: 87/100 (reviewer agent)
- Fixed bare exception handling with specific types
- Fixed type hint errors (any → Any)
- Fixed unused variable warnings from ruff
- All manual tests pass via CLI interfaces

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

---------

Co-authored-by: Ubuntu <azureuser@amplihack2.yb0a3bvkdghunmsjr4s3fnfhra.phxx.internal.cloudapp.net>
Co-authored-by: Claude <noreply@anthropic.com>

Author: 3 people

.claude/commands/amplihack/transcripts.md

@@ -49,9 +49,80 @@ Shows the original request and requirements for a session.
 
 - `session_id`: Target session (default: current)
 
+## What To Do
+
+When this command is invoked, use the `transcript_manager` tool to handle the request:
+
+### List Sessions (No Args or "list")
+
+```python
+from transcript_manager import TranscriptManager, list_transcripts, get_transcript_summary
+
+# List all available sessions
+sessions = list_transcripts()
+
+# Display formatted list with summaries
+manager = TranscriptManager()
+for i, session_id in enumerate(sessions[:10], 1):  # Show latest 10
+    summary = get_transcript_summary(session_id)
+    print(manager.format_summary_display(summary, i))
+```
+
+### Restore Latest Session ("latest")
+
+```python
+from transcript_manager import list_transcripts, restore_transcript, TranscriptManager
+
+sessions = list_transcripts()
+if sessions:
+    context = restore_transcript(sessions[0])
+    manager = TranscriptManager()
+    print(manager.format_context_display(context))
+```
+
+### Restore Specific Session (<session_id>)
+
+```python
+from transcript_manager import restore_transcript, TranscriptManager
+
+context = restore_transcript(session_id)
+manager = TranscriptManager()
+print(manager.format_context_display(context))
+# Display:
+# - Original user request
+# - Conversation summary
+# - Full transcript location
+# - Compaction events (if any)
+```
+
+### Save Checkpoint ("save")
+
+```python
+from transcript_manager import save_checkpoint, TranscriptManager
+from datetime import datetime
+
+session_id = save_checkpoint()
+manager = TranscriptManager()
+checkpoint_count = manager.get_checkpoint_count(session_id)
+
+# Display success message with:
+# - Session ID
+# - Checkpoint location
+# - Checkpoint number
+# - Timestamp
+print(f"✅ Session checkpoint created!")
+print(f"📄 Session ID: {session_id}")
+print(f"🔖 Checkpoint #{checkpoint_count}")
+print(f"⏰ Timestamp: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")
+```
+
 ## Implementation
 
-The actual implementation is in `/Users/ryan/src/hackathon/MicrosoftHackathon2025-AgenticCoding/.claude/commands/transcripts.py`
+All transcript functionality is provided by the `transcript_manager` tool:
+
+**Tool**: `.claude/tools/amplihack/transcript_manager.py`
+
+This command provides instructions on how to use the tool. No Python code is executed directly by this command - Claude interprets the instructions and calls the tool.
 
 ## Examples
 
@@ -130,4 +201,19 @@ Found 2 matches across sessions
 4. **Searchable History**: Find past decisions and implementations
 5. **Compaction Safe**: Automatic export before context loss
 
-This command provides amplihack-style transcript management, ensuring that conversation context and original requirements are never lost, even during context compaction events.
+## Philosophy Alignment
+
+- **Ruthless Simplicity**: Command is markdown instructions only, no Python code
+- **Separation of Concerns**: Command provides instructions, tool implements business logic
+- **Zero-BS**: All tool functions work completely, no stubs or placeholders
+- **Brick Philosophy**: transcript_manager.py is a self-contained, regeneratable brick
+
+## Resources
+
+- **Tool**: `.claude/tools/amplihack/transcript_manager.py` (business logic)
+- **Command**: `.claude/commands/amplihack/transcripts.md` (instructions only)
+- **PreCompact Hook**: `.claude/tools/amplihack/hooks/pre_compact.py` (automatic export)
+
+This command provides amplihack-style transcript management through a clean, reusable tool. The tool can be called from commands, skills, and hooks. It ensures that conversation context and original requirements are never lost, even during context compaction events.
+
+**Key Takeaway**: Business logic lives in `transcript_manager.py`, this command just tells you how to use it.