⚠️ MANDATORY: Use Task Tool (Sub-Agent)

NEVER call memory MCP tools directly! Use Task tool with subagent_type: "general-purpose" to keep main context clean.

---

CRITICAL: When NOT to Search Memory

Skip memory search for obvious tasks - killing processes, starting servers, basic file operations, standard workflows.

Only search for hard problems - non-obvious bugs, complex architectures, performance issues, unfamiliar domains.

Rule: If basic knowledge suffices, skip memory. Memory is for hard-won lessons.

---

Embedded Role Configuration

# Embedded configuration - no external files needed
role_collections:
  global:
    universal:
      name: "universal-patterns"
      description: "Search here for cross-domain patterns"
      query_hints: ["general", "architecture", "debugging", "performance"]

    backend:
      name: "backend-patterns"
      description: "Backend engineering patterns"
      query_hints: ["api", "database", "auth", "server", "microservices"]

    frontend:
      name: "frontend-patterns"
      description: "Frontend engineering patterns"
      query_hints: ["react", "vue", "component", "ui", "state"]

    quant:
      name: "quant-patterns"
      description: "Quantitative finance patterns"
      query_hints: ["trading", "backtest", "risk", "portfolio"]

    devops:
      name: "devops-patterns"
      description: "DevOps and infrastructure patterns"
      query_hints: ["docker", "kubernetes", "ci-cd", "terraform"]

    ai:
      name: "ai-patterns"
      description: "AI and machine learning patterns"
      query_hints: ["model", "training", "neural", "llm", "embedding"]

    security:
      name: "security-patterns"
      description: "Security engineering patterns"
      query_hints: ["vulnerability", "encryption", "auth", "pentest"]

    mobile:
      name: "mobile-patterns"
      description: "Mobile development patterns"
      query_hints: ["ios", "android", "react-native", "flutter"]

    pm:
      name: "pm-patterns"
      description: "Project management and coordination patterns"
      query_hints: ["coordination", "delegation", "team", "sprint", "planning", "reporting"]

# Role detection from task context
role_detection:
  patterns:
    backend: "api|endpoint|database|server|auth|rest|graphql"
    frontend: "react|vue|component|ui|dom|css|state"
    quant: "trading|backtest|portfolio|risk|market"
    devops: "deploy|docker|kubernetes|ci|cd"
    ai: "model|training|neural|embedding|llm"
    security: "vulnerability|encryption|pentest|jwt"
    mobile: "ios|android|native|flutter|swift"
    pm: "project|coordination|delegation|team|sprint|phase|reporting|stakeholder"

  multi_role_strategy: "search_all"  # When multiple roles detected
  default_role: "universal"          # When no clear role

You can create new role if you think it worth it. But be EXTREMELY CONSERVATIVE when creating new roles - when you create a new one, add it in this very doc (~/.claude/skills/coder-memory-recall/SKILL.md and ~/.claude/skills/coder-memory-store/SKILL.md).

PHASE 1: Intelligent Query Construction

Note: Claude Code automatically determines relevant roles from task context. No explicit role detection logic needed - Claude is smart enough to select appropriate roles when calling MCP tools.

Query Building

Build semantic query (2-3 sentences) capturing:

1. What is the problem/goal?
2. What is the technical context?
3. What outcome is desired?

MCP Server Tools

CRITICAL: Use tools from the memory MCP server:

• search_memory - Search and get previews
• get_memory - Get full content by ID
• batch_get_memories - Get multiple full contents
• store_memory - Store new memory
• update_memory - Update existing memory
• delete_memory - Delete memory
• list_collections - List all collections

PHASE 2: Two-Stage Retrieval

Stage 1: Search for Previews (Cast Wide Net)

Use search_memory tool (from memory MCP server) with the query and correct memory_level (global, project, etc.), default: memory_level="global". Claude Code determines relevant roles automatically. Default limit is 20 previews.

Stage 2: Analyze Previews (Intelligence Over Thresholds)

Analyze each preview:

• Does title match the problem domain?
• Does description indicate relevant solution?
• Do tags align with task?
• Is memory type appropriate? (episodic for debugging, procedural for workflows, semantic for principles)

Select 3-5 most relevant based on your judgement.

Stage 3: Retrieve Full Content

Use batch_get_memories tool (from memory MCP server) with the selected doc_ids and memory_level="global". This retrieves full content for 3-5 most relevant memories.

PHASE 3: Present Results

Format for Claude to consume: Key: Let Claude read and decide what to use. Don't force-fit patterns.

---

Tool Usage

See top of this document - MUST use Task tool (sub-agent) to avoid context pollution.