Get Pattern - Retrieve Application Knowledge

What This Skill Does

Retrieves established application patterns (architecture, procedures, conventions) for the Neural Data Platform using AgentDB's semantic vector search. Patterns are found by meaning, not just exact key match.

Use this BEFORE implementing anything to ensure you follow project standards.

CRITICAL: What This Skill IS and IS NOT For

USE FOR (Application Knowledge)

DO NOT USE FOR (Transient/Swarm Memory)

Patterns are PERMANENT application knowledge, not transient swarm state.

The Pattern Workflow

1. BEFORE work:  get-pattern  → Retrieve relevant patterns (THIS SKILL)
2. DURING work:  Apply the pattern, note what works
3. AFTER work:   reflexion    → Did get-pattern results help?
                 save-pattern → Store NEW discoveries (if any)

Quick Usage

Method 1: Semantic Search (Recommended)

Find patterns by describing what you're trying to do:

npx agentdb query --query "how to add a new data stream" --k 5 --min-confidence 0.7

Examples:

# Find architecture patterns
npx agentdb query --query "data pipeline architecture" --k 3

# Find deployment procedures
npx agentdb query --query "deploy to raspberry pi" --domain "deployment" --k 3

# Find naming conventions
npx agentdb query --query "naming conventions for streams and fields" --k 2

# Get synthesized context with patterns and insights
npx agentdb query --query "domain adapter pattern" --k 5 --synthesize-context

Method 2: Category-Filtered Search

When you know the domain or need filtered results:

# Filter by domain
npx agentdb query --query "add new data source" --domain "development" --k 3

# Use metadata filters (MongoDB-style)
npx agentdb query --query "add new data source" --filters '{"metadata.category":"procedure"}' --k 3

Method 3: Find Similar Past Experiences

See how similar tasks were handled before:

# Retrieve successful episodes
npx agentdb reflexion retrieve "adding a new MQTT data source" --k 5 --only-successes

# Get synthesized context from past experiences
npx agentdb reflexion retrieve "MQTT data source" --k 5 --synthesize-context

Pattern Categories

CRITICAL: Record Pattern Usage

After using a pattern, always use the reflexion skill to record whether it helped:

# Pattern worked well
npx agentdb reflexion store "ndp-patterns" \
  "Used [pattern-name] for [what you did]" \
  1.0 true \
  "Pattern was accurate and complete"

# Pattern needed fixes
npx agentdb reflexion store "ndp-patterns" \
  "Used [pattern-name] but needed adjustment" \
  0.6 true \
  "Pattern missing [gap] - should update via save-pattern"

Without feedback, the system can't learn which patterns actually work.

Key Architecture Documents (File References)

When patterns reference files, check these locations:

Common Pattern Lookups

Adding a New Data Stream

npx agentdb query --query "add new data stream to the platform" --k 3

Understanding the Architecture

npx agentdb query --query "system architecture and component relationships" --k 5 --synthesize-context

Naming Conventions

npx agentdb query --query "naming conventions for streams fields and modules" --k 3

Creating Grafana Dashboards

npx agentdb query --query "grafana dashboard creation with DuckDB" --domain "procedures" --k 3

Check Pattern Health

See which patterns exist and their success rates:

npx agentdb db stats

Returns: total patterns, average success rates, top task types, patterns needing review.

If Pattern Not Found

If no relevant pattern exists:

1. Implement the solution
2. If the approach is reusable, use save-pattern skill to store it
3. Future agents will benefit from your discovery

Related Skills

• save-pattern - Store NEW patterns for architecture, procedures, conventions
• reflexion - Record feedback on whether get-pattern results helped (REQUIRED after using patterns)

NOT related to:

• Swarm coordination (use claude-flow tools)
• Transient task memory (use claude-flow memory with TTL)
• Agent state management (use claude-flow tools)