Lev Workshop

Lev Workshop

The workshop is where ideas become capabilities. It's the factory floor for building hooks, plugins, and integrations that extend lev's functionality.

Workshop Lifecycle

┌─────────────┐    ┌─────────────┐    ┌─────────────┐    ┌─────────────┐
│   INTAKE    │ →  │  ANALYSIS   │ →  │     POC     │ →  │    POLY     │
│             │    │             │    │             │    │ INTEGRATION │
│ Raw idea or │    │ Understand  │    │ Build       │    │ Production  │
│ requirement │    │ scope &     │    │ working     │    │ deployment  │
│             │    │ approach    │    │ prototype   │    │ & routing   │
└─────────────┘    └─────────────┘    └─────────────┘    └─────────────┘
     │                   │                   │                   │
     ▼                   ▼                   ▼                   ▼
 config_loader     dependency_checker   checkpoint_manager   quality_gate
 error_handler     progress_reporter    error_handler        success_handler
                   quality_gate         quality_gate         escalation_handler

When to Use

Phase Details

Phase 1: Intake

Purpose: Capture raw ideas, requirements, or problems that need solving.

Input: Natural language description, error message, feature request, or observed pattern.

Process:

1. config_loader - Load relevant domain context
2. error_handler - If intake is from error, parse and classify
3. Classify intake type:
   • hook - Lifecycle hook meta-prompt
   • plugin - External integration
   • pattern - Reusable workflow
   • fix - Bug fix or improvement

Output: Workshop item in ~/lev/workshop/intake/

# ~/lev/workshop/intake/item-001.yaml
id: ws-001
type: hook
title: "Retry with circuit breaker"
source: "Observed repeated failures in CB3 scraping"
created: "2026-01-13T10:00:00Z"
status: intake
raw_input: |
  CB3 keeps hitting rate limits and retrying forever.
  Need a circuit breaker pattern.

Phase 2: Analysis

Purpose: Understand scope, dependencies, and approach before building.

Process:

1. dependency_checker - What does this need?
2. progress_reporter - Track analysis progress
3. Research existing implementations
4. Define success criteria
5. quality_gate - Is analysis complete enough to proceed?

Questions to answer:

• What existing hooks/plugins does this interact with?
• What's the minimal viable implementation?
• What are the edge cases?
• How will we test it?

Output: Analysis document added to workshop item

# Updated item-001.yaml
analysis:
  dependencies:
    - error_handler (hook)
    - checkpoint_manager (hook)
  approach: |
    1. Track failure count per operation
    2. Open circuit after N failures
    3. Half-open state for testing
    4. Auto-close after cooldown
  success_criteria:
    - Circuit opens after 3 failures
    - Half-open after 30s
    - Full close after 3 successes
  estimated_complexity: medium
  analysis_complete: true

Phase 3: POC (Proof of Concept)

Purpose: Build a working prototype to validate the approach.

Process:

1. checkpoint_manager - Save state frequently during development
2. Build minimal implementation
3. error_handler - Handle development errors
4. Test against success criteria
5. quality_gate - Does POC meet criteria?

Location: ~/lev/workshop/poc/<item-id>/

POC requirements:

• Must run standalone
• Must have basic tests
• Must document usage
• Must log decisions

Output: Working prototype with test results

# Updated item-001.yaml
poc:
  location: ~/lev/workshop/poc/ws-001/
  files:
    - circuit_breaker.py
    - test_circuit_breaker.py
    - README.md
  test_results:
    passed: 8
    failed: 0
  demo_command: "python circuit_breaker.py --demo"
  poc_complete: true

Phase 4: Poly Integration

Purpose: Promote POC to production across multiple integration targets.

Process:

1. quality_gate - Final validation before integration
2. Identify integration targets (poly = multiple destinations)
3. success_handler - Route to each target
4. escalation_handler - Handle integration failures

Integration targets:

• lev-lifecycle - New hook added to hooks.md
• claude-agent-sdk - Pattern added to patterns.md
• Project-specific - CB3, TimeTravel, josh-imessage, etc.
• System - ~/lev/plugins/ and project hooks for automated workflows

Poly routing abstraction:

# Integration targets can register as recipients
integration_targets:
  lev_hooks:
    type: hook
    location: ~/.claude/skills/lev-lifecycle/references/hooks.md
    format: yaml_block

  agent_sdk_patterns:
    type: pattern
    location: ~/.claude/skills/claude-agent-sdk/references/patterns.md
    format: python_code

  cb3_tools:
    type: tool
    location: ~/cb3/src/tools/
    format: python_module

  timetravel_adapters:
    type: adapter
    location: ~/lev/research/timetravel/src/adapters/
    format: typescript_module

Output: Integrated capability with routing info

# Updated item-001.yaml
integration:
  status: complete
  integrated_to:
    - target: lev_hooks
      location: ~/.claude/skills/lev-lifecycle/references/hooks.md
      section: "Hook 11: Circuit Breaker"
    - target: cb3_tools
      location: ~/cb3/src/tools/circuit_breaker.py
  routing_rules:
    - pattern: "error_handler returns retry && retry_count >= 3"
      route_to: circuit_breaker
  post_integration_hooks:
    - update_docs
    - notify_user

Workshop Item Schema

id: ws-XXX
type: hook | plugin | pattern | fix
title: string
source: string  # where this came from
created: timestamp
status: intake | analysis | poc | integrating | complete | archived

# Phase outputs
raw_input: string
analysis:
  dependencies: array
  approach: string
  success_criteria: array
  estimated_complexity: low | medium | high
  analysis_complete: boolean

poc:
  location: path
  files: array
  test_results: object
  demo_command: string
  poc_complete: boolean

integration:
  status: pending | in_progress | complete | failed
  integrated_to: array
  routing_rules: array
  post_integration_hooks: array

# Metadata
bd_issue: string  # linked bd bead
related_items: array
tags: array

Process Tracking

Workshop items are tracked in bd for visibility across sessions:

# Create bd issue for workshop item
bd create "Workshop: Circuit Breaker Hook" -t feature -p 2 \
  -d "Build circuit breaker pattern for retry logic"

# Link workshop item to bd
# In item-001.yaml:
# bd_issue: ws-circuit-breaker-001

# Query workshop status
bd list --label workshop --status open

CLI Commands

# Start new workshop item
lev workshop intake "Need circuit breaker for retries"

# Check workshop status
lev workshop status

# List items by phase
lev workshop list --phase poc

# Promote to next phase
lev workshop promote ws-001 --to analysis
lev workshop promote ws-001 --to poc
lev workshop promote ws-001 --to integration

# Run POC tests
lev workshop test ws-001

# Integrate to target
lev workshop integrate ws-001 --target lev_hooks

# Archive completed item
lev workshop archive ws-001

Example: Building a New Hook

# 1. Intake
lev workshop intake "Error handler should support circuit breaker pattern"
# → Creates ws-001 in intake phase

# 2. Analysis
lev workshop analyze ws-001
# Agent asks questions, researches existing hooks
# → Adds analysis section, promotes to analysis phase

# 3. POC
lev workshop poc ws-001
# Agent builds prototype in poc/ws-001/
# → Creates files, runs tests, promotes to poc phase

# 4. Integration
lev workshop integrate ws-001 --target lev_hooks
# Agent adds to hooks.md, updates references
# → Routes to all configured targets

# 5. Complete
lev workshop complete ws-001
# Archives item, updates bd issue

Integration with Claude Agent SDK

Workshop items that involve LLM decision points use the Claude Agent SDK:

from claude_agent_sdk import query, ClaudeAgentOptions

async def analyze_workshop_item(item_id: str):
    """Use agent to analyze workshop item requirements"""
    async for msg in query(
        prompt=f"""
        Analyze workshop item {item_id} for implementation:
        1. Read the raw input
        2. Identify dependencies on existing hooks/plugins
        3. Propose minimal implementation approach
        4. Define success criteria
        5. Estimate complexity
        """,
        options=ClaudeAgentOptions(
            allowed_tools=["Read", "Grep", "Glob"],
            permission_mode="bypassPermissions"
        )
    ):
        if hasattr(msg, "result"):
            return msg.result

Hooks Used Per Phase

References

• ~/.claude/skills/lev-lifecycle/ - Parent lifecycle skill
• ~/.claude/skills/lev-lifecycle/references/hooks.md - Hook meta-prompts
• ~/.claude/skills/claude-agent-sdk/ - Agent patterns for LLM integration
• ~/lev/workshop/ - Workshop storage

Relates

Master Router

• Lev Master Router (lev/SKILL.md) - Routes all lev-* skills Parent skill that dispatches to this skill based on keywords/context

Technique Map

• Role definition - Clarifies operating scope and prevents ambiguous execution.
• Context enrichment - Captures required inputs before actions.
• Output structuring - Standardizes deliverables for consistent reuse.
• Step-by-step workflow - Reduces errors by making execution order explicit.
• Edge-case handling - Documents safe fallbacks when assumptions fail.

Technique Notes

These techniques improve reliability by making intent, inputs, outputs, and fallback paths explicit. Keep this section concise and additive so existing domain guidance remains primary.

Prompt Architect Overlay

Role Definition

You are the prompt-architect-enhanced specialist for lev-workshop, responsible for deterministic execution of this skill's guidance while preserving existing workflow and constraints.

Input Contract

• Required: clear user intent and relevant context for this skill.
• Preferred: repository/project constraints, existing artifacts, and success criteria.
• If context is missing, ask focused questions before proceeding.

Output Contract

• Provide structured, actionable outputs aligned to this skill's existing format.
• Include assumptions and next steps when appropriate.
• Preserve compatibility with existing sections and related skills.

Edge Cases & Fallbacks

• If prerequisites are missing, provide a minimal safe path and request missing inputs.
• If scope is ambiguous, narrow to the highest-confidence sub-task.
• If a requested action conflicts with existing constraints, explain and offer compliant alternatives.