Retrieve universal coding patterns from vector database using true two-stage retrieval. Auto-invokes before complex tasks or when user says "--recall". Searches relevant role collections based on task context.
Retrieves relevant coding patterns from vector database using intelligent two-stage search.
npx claudepluginhub hongbietcode/synthetic-claudeThis skill inherits all available tools. When active, it can use any tool Claude has access to.
NEVER call memory MCP tools directly! Use Task tool with subagent_type: "general-purpose" to keep main context clean.
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 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).
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.
Build semantic query (2-3 sentences) capturing:
CRITICAL: Use tools from the memory MCP server:
search_memory - Search and get previewsget_memory - Get full content by IDbatch_get_memories - Get multiple full contentsstore_memory - Store new memoryupdate_memory - Update existing memorydelete_memory - Delete memorylist_collections - List all collectionsUse 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.
Analyze each preview:
Select 3-5 most relevant based on your judgement.
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.
Format for Claude to consume: Key: Let Claude read and decide what to use. Don't force-fit patterns.
See top of this document - MUST use Task tool (sub-agent) to avoid context pollution.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Applies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.