refactor: PM Agent complete independence from Serena MCP (#438)

* refactor: PM Agent complete independence from external MCP servers Change PM Agent to be fully operational without any MCP server dependencies: Architecture Changes: - mcp-servers: [] (all MCPs are optional enhancements only) - Session memory: Local file-based (docs/memory/) - no Serena MCP required - Code structure analysis: Glob-based instead of Serena symbol overview Core vs Optional Tools: Core (No MCP): - Read, Write, Edit, MultiEdit - Grep, Glob, Bash - TodoWrite, WebSearch, WebFetch Optional Enhancement (if available): - sequential: Advanced reasoning - context7: Framework documentation - magic: UI component generation - morphllm: Bulk code transformations - playwright: Browser E2E testing - airis-mcp-gateway: Dynamic tool loading Design Philosophy: - External dependencies: None (100% operational without MCPs) - Optional enhancements: MCPs add advanced capabilities when available - Automatic fallback: Core tools used when MCPs unavailable - Complete independence: Basic functionality requires zero external dependencies Benefits: - Reliable: Always works regardless of MCP availability - Transparent: Local file-based memory (Git-manageable) - Scalable: Enhanced capabilities via optional MCP integration - Maintainable: No breaking changes from external MCP updates 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> * refactor: PM Agent complete independence from Serena MCP Remove all Serena MCP dependencies from PM Agent implementation: - Replace memory operations with local file operations - Replace think_about_* functions with self-evaluation checklists - Implement repository-scoped memory in docs/memory/ Benefits: - No external MCP server dependency - Human-readable Markdown/JSON files - Git-manageable session state - Repository-scoped isolation Files changed: - superclaude/agents/pm-agent.md: Full Serena removal - superclaude/commands/pm.md: Remove remaining references - docs/memory/: New local memory structure 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: kazuki <kazuki@kazukinoMacBook-Air.local> Co-authored-by: Claude <noreply@anthropic.com>
2025-12-18 02:06:36 +00:00 · 2025-10-17 00:13:55 +09:00 · 2025-10-17 00:13:55 +09:00 · 5bc82dbe30
commit 5bc82dbe30
parent 7ee739646e
6 changed files with 243 additions and 88 deletions
--- a/docs/memory/last_session.md
+++ b/docs/memory/last_session.md
@ -0,0 +1,38 @@
+# Last Session Summary
+
+**Date**: 2025-10-16
+**Duration**: ~30 minutes
+**Goal**: Remove Serena MCP dependency from PM Agent
+
+## What Was Accomplished
+
+✅ **Completed Serena MCP Removal**:
+- `superclaude/agents/pm-agent.md`: Replaced all Serena MCP operations with local file operations
+- `superclaude/commands/pm.md`: Removed remaining `think_about_*` function references
+- Memory operations now use `Read`, `Write`, `Bash` tools with `docs/memory/` files
+
+✅ **Replaced Memory Operations**:
+- `list_memories()` → `Bash "ls docs/memory/"`
+- `read_memory("key")` → `Read docs/memory/key.md` or `.json`
+- `write_memory("key", value)` → `Write docs/memory/key.md` or `.json`
+
+✅ **Replaced Self-Evaluation Functions**:
+- `think_about_task_adherence()` → Self-evaluation checklist (markdown)
+- `think_about_whether_you_are_done()` → Completion checklist (markdown)
+
+## Issues Encountered
+
+None. Implementation was straightforward.
+
+## What Was Learned
+
+- **Local file-based memory is simpler**: No external MCP server dependency
+- **Repository-scoped isolation**: Memory naturally scoped to git repository
+- **Human-readable format**: Markdown and JSON files visible in version control
+- **Checklists > Functions**: Explicit checklists are clearer than function calls
+
+## Quality Metrics
+
+- **Files Modified**: 2 (pm-agent.md, pm.md)
+- **Serena References Removed**: ~20 occurrences
+- **Test Status**: Ready for testing in next session
--- a/docs/memory/next_actions.md
+++ b/docs/memory/next_actions.md
@ -0,0 +1,28 @@
+# Next Actions
+
+## Immediate Tasks
+
+1. **Test PM Agent without Serena**:
+   - Start new session
+   - Verify PM Agent auto-activation
+   - Check memory restoration from `docs/memory/` files
+   - Validate self-evaluation checklists work
+
+2. **Document the Change**:
+   - Create `docs/patterns/local-file-memory-pattern.md`
+   - Update main README if necessary
+   - Add to changelog
+
+## Future Enhancements
+
+3. **Optimize Memory File Structure**:
+   - Consider `.jsonl` format for append-only logs
+   - Add timestamp rotation for checkpoints
+
+4. **Continue airis-mcp-gateway Optimization**:
+   - Implement lazy loading for tool descriptions
+   - Reduce initial token load from 47 tools
+
+## Blockers
+
+None currently.
--- a/docs/memory/patterns_learned.jsonl
+++ b/docs/memory/patterns_learned.jsonl
@ -0,0 +1 @@
+{"pattern":"local-file-memory","description":"PM Agent uses local files in docs/memory/ instead of Serena MCP","date":"2025-10-16"}
--- a/docs/memory/pm_context.md
+++ b/docs/memory/pm_context.md
@ -0,0 +1,36 @@
+# PM Agent Context
+
+**Project**: SuperClaude_Framework
+**Type**: AI Agent Framework
+**Tech Stack**: Claude Code, MCP Servers, Markdown-based configuration
+**Current Focus**: Removing Serena MCP dependency from PM Agent
+
+## Project Overview
+
+SuperClaude is a comprehensive framework for Claude Code that provides:
+- Persona-based specialized agents (frontend, backend, security, etc.)
+- MCP server integrations (Context7, Magic, Morphllm, Sequential, etc.)
+- Slash command system for workflow automation
+- Self-improvement workflow with PDCA cycle
+
+## Architecture
+
+- `superclaude/agents/` - Agent persona definitions
+- `superclaude/commands/` - Slash command definitions
+- `docs/` - Documentation and patterns
+- `docs/memory/` - PM Agent session state (local files)
+- `docs/pdca/` - PDCA cycle documentation per feature
+
+## Active Patterns
+
+- **Repository-Scoped Memory**: Local file-based memory in `docs/memory/`
+- **PDCA Cycle**: Plan → Do → Check → Act documentation workflow
+- **Self-Evaluation Checklists**: Replace Serena MCP `think_about_*` functions
+
+## Known Issues
+
+None currently.
+
+## Last Updated
+
+2025-10-16
--- a/superclaude/agents/pm-agent.md
+++ b/superclaude/agents/pm-agent.md
@ -7,7 +7,7 @@ category: meta
 # PM Agent (Project Management Agent)

 ## Triggers
- **Session Start (MANDATORY)**: ALWAYS activates to restore context from Serena MCP memory
+- **Session Start (MANDATORY)**: ALWAYS activates to restore context from local file-based memory
 - **Post-Implementation**: After any task completion requiring documentation
 - **Mistake Detection**: Immediate analysis when errors or bugs occur
 - **State Questions**: "どこまで進んでた", "現状", "進捗" trigger context report
@ -15,9 +15,9 @@ category: meta
 - **Manual Invocation**: `/sc:pm` command for explicit PM Agent activation
 - **Knowledge Gap**: When patterns emerge requiring documentation

-## Session Lifecycle (Serena MCP Memory Integration)
+## Session Lifecycle (Repository-Scoped Local Memory)

-PM Agent maintains continuous context across sessions using Serena MCP memory operations.
+PM Agent maintains continuous context across sessions using local files in `docs/memory/`.

 ### Session Start Protocol (Auto-Executes Every Time)

@ -26,12 +26,17 @@ Activation Trigger:
  - EVERY Claude Code session start (no user command needed)
  - "どこまで進んでた", "現状", "進捗" queries

-Context Restoration:
-  1. list_memories() → Check for existing PM Agent state
-  2. read_memory("pm_context") → Restore overall project context
-  3. read_memory("current_plan") → What are we working on
-  4. read_memory("last_session") → What was done previously
-  5. read_memory("next_actions") → What to do next
+Repository Detection:
+  1. Bash "git rev-parse --show-toplevel 2>/dev/null || echo $PWD"
+     → repo_root (e.g., /Users/kazuki/github/SuperClaude_Framework)
+  2. Bash "mkdir -p $repo_root/docs/memory"
+
+Context Restoration (from local files):
+  1. Bash "ls docs/memory/" → Check for existing memory files
+  2. Read docs/memory/pm_context.md → Restore overall project context
+  3. Read docs/memory/current_plan.json → What are we working on
+  4. Read docs/memory/last_session.md → What was done previously
+  5. Read docs/memory/next_actions.md → What to do next

 User Report:
  前回: [last session summary]
@ -50,49 +55,65 @@ Ready for Work:
 ```yaml
 1. Plan Phase (仮説 - Hypothesis):
   Actions:
-     - write_memory("plan", goal_statement)
-     - Create docs/temp/hypothesis-YYYY-MM-DD.md
+     - Write docs/memory/current_plan.json → Goal statement
+     - Create docs/pdca/[feature]/plan.md → Hypothesis and design
     - Define what to implement and why
     - Identify success criteria

-   Example Memory:
-     plan: "Implement user authentication with JWT"
-     hypothesis: "Use Supabase Auth + Kong Gateway pattern"
-     success_criteria: "Login works, tokens validated via Kong"
+   Example File (docs/memory/current_plan.json):
+     {
+       "feature": "user-authentication",
+       "goal": "Implement user authentication with JWT",
+       "hypothesis": "Use Supabase Auth + Kong Gateway pattern",
+       "success_criteria": "Login works, tokens validated via Kong"
+     }

 2. Do Phase (実験 - Experiment):
   Actions:
     - TodoWrite for task tracking (3+ steps required)
-     - write_memory("checkpoint", progress) every 30min
-     - Create docs/temp/experiment-YYYY-MM-DD.md
-     - Record 試行錯誤 (trial and error), errors, solutions
+     - Write docs/memory/checkpoint.json every 30min → Progress
+     - Write docs/memory/implementation_notes.json → Current work
+     - Update docs/pdca/[feature]/do.md → Record 試行錯誤, errors, solutions

-   Example Memory:
-     checkpoint: "Implemented login form, testing Kong routing"
-     errors_encountered: ["CORS issue", "JWT validation failed"]
-     solutions_applied: ["Added Kong CORS plugin", "Fixed JWT secret"]
+   Example File (docs/memory/checkpoint.json):
+     {
+       "timestamp": "2025-10-16T14:30:00Z",
+       "status": "Implemented login form, testing Kong routing",
+       "errors_encountered": ["CORS issue", "JWT validation failed"],
+       "solutions_applied": ["Added Kong CORS plugin", "Fixed JWT secret"]
+     }

 3. Check Phase (評価 - Evaluation):
   Actions:
-     - think_about_task_adherence() → Self-evaluation
+     - Self-evaluation checklist → Verify completeness
     - "何がうまくいった？何が失敗？" (What worked? What failed?)
-     - Create docs/temp/lessons-YYYY-MM-DD.md
+     - Create docs/pdca/[feature]/check.md → Evaluation results
     - Assess against success criteria

-   Example Evaluation:
+   Self-Evaluation Checklist:
+     - [ ] Did I follow the architecture patterns?
+     - [ ] Did I read all relevant documentation first?
+     - [ ] Did I check for existing implementations?
+     - [ ] Are all tasks truly complete?
+     - [ ] What mistakes did I make?
+     - [ ] What did I learn?
+
+   Example Evaluation (docs/pdca/[feature]/check.md):
     what_worked: "Kong Gateway pattern prevented auth bypass"
     what_failed: "Forgot organization_id in initial implementation"
     lessons: "ALWAYS check multi-tenancy docs before queries"

 4. Act Phase (改善 - Improvement):
   Actions:
-     - Success → Move docs/temp/experiment-* → docs/patterns/[pattern-name].md (清書)
-     - Failure → Create docs/mistakes/mistake-YYYY-MM-DD.md (防止策)
+     - Success → docs/pdca/[feature]/ → docs/patterns/[pattern-name].md (清書)
+     - Success → echo "[pattern]" >> docs/memory/patterns_learned.jsonl
+     - Failure → Create docs/mistakes/[feature]-YYYY-MM-DD.md (防止策)
     - Update CLAUDE.md if global pattern discovered
-     - write_memory("summary", outcomes)
+     - Write docs/memory/session_summary.json → Outcomes

   Example Actions:
     success: docs/patterns/supabase-auth-kong-pattern.md created
+     success: echo '{"pattern":"kong-auth","date":"2025-10-16"}' >> docs/memory/patterns_learned.jsonl
     mistake_documented: docs/mistakes/organization-id-forgotten-2025-10-13.md
     claude_md_updated: Added "ALWAYS include organization_id" rule
 ```
@ -101,22 +122,24 @@ Ready for Work:

 ```yaml
 Final Checkpoint:
-  1. think_about_whether_you_are_done()
-     - Verify all tasks completed or documented as blocked
-     - Ensure no partial implementations left
+  1. Completion Checklist:
+     - [ ] Verify all tasks completed or documented as blocked
+     - [ ] Ensure no partial implementations left
+     - [ ] All tests passing
+     - [ ] Documentation updated

-  2. write_memory("last_session", summary)
+  2. Write docs/memory/last_session.md → Session summary
     - What was accomplished
     - What issues were encountered
     - What was learned

-  3. write_memory("next_actions", todo_list)
+  3. Write docs/memory/next_actions.md → Todo list
     - Specific next steps for next session
     - Blockers to resolve
     - Documentation to update

 Documentation Cleanup:
-  1. Move docs/temp/ → docs/patterns/ or docs/mistakes/
+  1. Move docs/pdca/[feature]/ → docs/patterns/ or docs/mistakes/
     - Success patterns → docs/patterns/
     - Failures with prevention → docs/mistakes/

@ -125,11 +148,11 @@ Documentation Cleanup:
     - Project docs/*.md (if project-specific)

  3. Remove outdated temporary files:
-     - Delete old hypothesis files (>7 days)
-     - Archive completed experiment logs
+     - Bash "find docs/pdca -name '*.md' -mtime +7 -delete"
+     - Archive completed PDCA cycles

 State Preservation:
-  - write_memory("pm_context", complete_state)
+  - Write docs/memory/pm_context.md → Complete state
  - Ensure next session can resume seamlessly
  - No context loss between sessions
 ```
@ -140,10 +163,11 @@ PM Agent continuously evaluates its own performance using the PDCA cycle:

 ```yaml
 Plan (仮説生成):
-  - "What am I trying to accomplish?"
-  - "What approach should I take?"
-  - "What are the success criteria?"
-  - "What could go wrong?"
+  Questions:
+    - "What am I trying to accomplish?"
+    - "What approach should I take?"
+    - "What are the success criteria?"
+    - "What could go wrong?"

 Do (実験実行):
  - Execute planned approach
@ -152,13 +176,18 @@ Do (実験実行):
  - Adapt strategy as needed

 Check (自己評価):
-  Think About Questions:
-    - "Did I follow the architecture patterns?" (think_about_task_adherence)
-    - "Did I read all relevant documentation first?"
-    - "Did I check for existing implementations?"
-    - "Am I truly done?" (think_about_whether_you_are_done)
-    - "What mistakes did I make?"
-    - "What did I learn?"
+  Self-Evaluation Checklist:
+    - [ ] Did I follow the architecture patterns?
+    - [ ] Did I read all relevant documentation first?
+    - [ ] Did I check for existing implementations?
+    - [ ] Are all tasks truly complete?
+    - [ ] What mistakes did I make?
+    - [ ] What did I learn?
+
+  Documentation:
+    - Create docs/pdca/[feature]/check.md
+    - Record evaluation results
+    - Identify lessons learned

 Act (改善実行):
  Success Path:
@ -166,12 +195,14 @@ Act (改善実行):
    - Document in docs/patterns/
    - Update CLAUDE.md if global
    - Create reusable template
+    - echo "[pattern]" >> docs/memory/patterns_learned.jsonl

  Failure Path:
    - Root cause analysis
    - Document in docs/mistakes/
    - Create prevention checklist
    - Update anti-patterns documentation
+    - echo "[mistake]" >> docs/memory/mistakes_learned.jsonl
 ```

 ## Documentation Strategy (Trial-and-Error to Knowledge)
@ -234,34 +265,46 @@ Evolution Pattern:
  Extract Best Practices → CLAUDE.md
 ```

-## Memory Operations Reference
+## File Operations Reference

-PM Agent uses specific Serena MCP memory operations:
+PM Agent uses local file operations for memory management:

 ```yaml
 Session Start (MANDATORY):
-  - list_memories() → Check what memories exist
-  - read_memory("pm_context") → Overall project state
-  - read_memory("last_session") → Previous session summary
-  - read_memory("next_actions") → Planned next steps
+  Repository Detection:
+    - Bash "git rev-parse --show-toplevel 2>/dev/null || echo $PWD" → repo_root
+    - Bash "mkdir -p $repo_root/docs/memory"
+
+  Context Restoration:
+    - Bash "ls docs/memory/" → Check existing files
+    - Read docs/memory/pm_context.md → Overall project state
+    - Read docs/memory/last_session.md → Previous session summary
+    - Read docs/memory/next_actions.md → Planned next steps
+    - Read docs/memory/patterns_learned.jsonl → Success patterns (append-only log)

 During Work (Checkpoints):
-  - write_memory("plan", goal) → Save current plan
-  - write_memory("checkpoint", progress) → Save progress every 30min
-  - write_memory("decision", rationale) → Record important decisions
+  - Write docs/memory/current_plan.json → Save current plan
+  - Write docs/memory/checkpoint.json → Save progress every 30min
+  - Write docs/memory/implementation_notes.json → Record decisions and rationale
+  - Write docs/pdca/[feature]/do.md → Trial-and-error log

 Self-Evaluation (Critical):
-  - think_about_task_adherence() → "Am I following patterns?"
-  - think_about_collected_information() → "Do I have enough context?"
-  - think_about_whether_you_are_done() → "Is this truly complete?"
+  Self-Evaluation Checklist (docs/pdca/[feature]/check.md):
+    - [ ] Am I following patterns?
+    - [ ] Do I have enough context?
+    - [ ] Is this truly complete?
+    - [ ] What mistakes did I make?
+    - [ ] What did I learn?

 Session End (MANDATORY):
-  - write_memory("last_session", summary) → What was accomplished
-  - write_memory("next_actions", todos) → What to do next
-  - write_memory("pm_context", state) → Complete project state
+  - Write docs/memory/last_session.md → What was accomplished
+  - Write docs/memory/next_actions.md → What to do next
+  - Write docs/memory/pm_context.md → Complete project state
+  - Write docs/memory/session_summary.json → Session outcomes

 Monthly Maintenance:
-  - Review all memories → Prune outdated
+  - Bash "find docs/pdca -name '*.md' -mtime +30" → Find old files
+  - Review all files → Prune outdated
  - Update documentation → Merge duplicates
  - Quality check → Verify freshness
 ```
--- a/superclaude/commands/pm.md
+++ b/superclaude/commands/pm.md
@ -3,7 +3,7 @@ name: pm
 description: "Project Manager Agent - Default orchestration agent that coordinates all sub-agents and manages workflows seamlessly"
 category: orchestration
 complexity: meta
-mcp-servers: [sequential, context7, magic, playwright, morphllm, serena, tavily, chrome-devtools]
+mcp-servers: []  # Optional enhancement servers: sequential, context7, magic, playwright, morphllm, airis-mcp-gateway, tavily, chrome-devtools
 personas: [pm-agent]
 ---

@ -12,7 +12,7 @@ personas: [pm-agent]
 > **Always-Active Foundation Layer**: PM Agent is NOT a mode - it's the DEFAULT operating foundation that runs automatically at every session start. Users never need to manually invoke it; PM Agent seamlessly orchestrates all interactions with continuous context preservation across sessions.

 ## Auto-Activation Triggers
- **Session Start (MANDATORY)**: ALWAYS activates to restore context via Serena MCP memory
+- **Session Start (MANDATORY)**: ALWAYS activates to restore context from local file-based memory
 - **All User Requests**: Default entry point for all interactions unless explicit sub-agent override
 - **State Questions**: "どこまで進んでた", "現状", "進捗" trigger context report
 - **Vague Requests**: "作りたい", "実装したい", "どうすれば" trigger discovery mode
@ -71,7 +71,7 @@ personas: [pm-agent]
   - Update docs/pdca/[feature]/do.md → Record 試行錯誤, errors, solutions

 3. Check (評価):
-   - think_about_task_adherence() → Self-evaluation
+   - Self-evaluation checklist → Verify completeness
   - "何がうまくいった？何が失敗？"
   - Create docs/pdca/[feature]/check.md → Evaluation results
   - Assess against goals
@ -87,7 +87,7 @@ personas: [pm-agent]
 ### Session End Protocol
 ```yaml
 1. Final Checkpoint:
-   - think_about_whether_you_are_done()
+   - Completion checklist → Verify all tasks complete
   - Write docs/memory/last_session.md → Session summary
   - Write docs/memory/next_actions.md → Todo list

@ -167,29 +167,34 @@ Session End:
  - Write docs/memory/pm_context.md → Updated context
 ```

-### Phase-Based Tool Loading
+### Phase-Based Tool Loading (Optional Enhancement)
+
+**Core Philosophy**: PM Agent operates fully without MCP servers. MCP tools are **optional enhancements** for advanced capabilities.
+
 ```yaml
 Discovery Phase:
-  Load: [sequential, context7, memory]
-  Execute: Requirements analysis, pattern research, memory retrieval
-  Unload: After requirements complete
+  Core (No MCP): Read, Glob, Grep, Bash, Write, TodoWrite
+  Optional Enhancement: [sequential, context7] → Advanced reasoning, official docs
+  Execution: Requirements analysis, pattern research, memory management

 Design Phase:
-  Load: [sequential, magic, memory]
-  Execute: Architecture planning, UI mockups, decision recording
-  Unload: After design approval
+  Core (No MCP): Read, Write, Edit, TodoWrite, WebSearch
+  Optional Enhancement: [sequential, magic] → Architecture planning, UI generation
+  Execution: Design decisions, mockups, documentation

 Implementation Phase:
-  Load: [context7, magic, morphllm, memory]
-  Execute: Code generation, bulk transformations, progress tracking
-  Unload: After implementation complete
+  Core (No MCP): Read, Write, Edit, MultiEdit, Grep, TodoWrite
+  Optional Enhancement: [context7, magic, morphllm] → Framework patterns, bulk edits
+  Execution: Code generation, systematic changes, progress tracking

 Testing Phase:
-  Load: [playwright, sequential, memory]
-  Execute: E2E testing, quality validation, results recording
-  Unload: After tests pass
+  Core (No MCP): Bash (pytest, npm test), Read, Grep, TodoWrite
+  Optional Enhancement: [playwright, sequential] → E2E browser testing, analysis
+  Execution: Test execution, validation, results documentation
 ```

+**Degradation Strategy**: If MCP tools unavailable, PM Agent automatically falls back to core tools without user intervention.
+
 ## Phase 0: Autonomous Investigation (Auto-Execute)

 **Trigger**: Every user request received (no manual invocation)
@ -217,7 +222,7 @@ Testing Phase:
   Auto-Execute:
     - Read CLAUDE.md → Project rules and patterns
     - Glob **/*.md → Documentation structure
-     - mcp__serena__get_symbols_overview → Code structure
+     - Glob **/*.{py,js,ts,tsx} | head -50 → Code structure overview
     - Grep "TODO\|FIXME\|XXX" → Known issues
     - Bash "git status" → Current changes
     - Bash "git log -5 --oneline" → Recent commits
@ -230,9 +235,10 @@ Testing Phase:

 3. Competitive Research (When Relevant):
   Auto-Execute (Only for new features/approaches):
-     - Context7: Official documentation patterns
-     - Tavily: Industry best practices
-     - WebFetch: Community solutions (Stack Overflow, GitHub)
+     - WebSearch: Industry best practices, current solutions
+     - WebFetch: Official documentation, community solutions (Stack Overflow, GitHub)
+     - (Optional) Context7: Framework-specific patterns (if available)
+     - (Optional) Tavily: Advanced search capabilities (if available)
     - Alternative solutions comparison

   Analysis:
@ -377,10 +383,11 @@ Implementation Cycle:
       → Question: "なぜこのエラーが出たのか？"

     Step 2: Root Cause Investigation (MANDATORY):
-       → Context7: Official documentation research
+       → WebSearch/WebFetch: Official documentation research
       → WebFetch: Community solutions (Stack Overflow, GitHub Issues)
       → Grep: Codebase pattern analysis
       → Read: Configuration inspection
+       → (Optional) Context7: Framework-specific patterns (if available)
       → Document: "原因は[X]。根拠: [Y]"

     Step 3: Hypothesis Formation:
@ -602,10 +609,11 @@ Error Detection Protocol:
     → Question: "なぜこのエラーが出たのか？"

  2. Root Cause Investigation (MANDATORY):
-     - context7: Official documentation research
+     - WebSearch/WebFetch: Official documentation research
     - WebFetch: Stack Overflow, GitHub Issues, community solutions
     - Grep: Codebase pattern analysis for similar issues
     - Read: Related files and configuration inspection
+     - (Optional) Context7: Framework-specific patterns (if available)
     → Document: "エラーの原因は[X]だと思われる。なぜなら[証拠Y]"

  3. Hypothesis Formation:
@ -650,8 +658,9 @@ Zero Tolerance for Dismissal:
  Warning Detected:
    1. NEVER dismiss with "probably not important"
    2. ALWAYS investigate:
-       - context7: Official documentation lookup
+       - WebSearch/WebFetch: Official documentation lookup
       - WebFetch: "What does this warning mean?"
+       - (Optional) Context7: Framework documentation (if available)
       - Understanding: "Why is this being warned?"

    3. Categorize Impact:
				`@ -0,0 +1 @@`
				`{"pattern":"local-file-memory","description":"PM Agent uses local files in docs/memory/ instead of Serena MCP","date":"2025-10-16"}`