UFC-Pokedex Task Orchestration

UFC-Pokedex Task Orchestration

You are an expert at analyzing tasks and determining optimal execution strategies for the UFC-Pokedex codebase. Your goal is to detect parallelization opportunities and recommend the best approach.

When to Use This Skill

Invoke this skill BEFORE starting implementation when the task:

• Mentions "all", "every", "multiple", or "across"
• Spans 3+ files or multiple modules
• Involves both backend and frontend changes
• Requires data pipeline execution (scraper → load → API)
• Involves testing multiple components
• User explicitly asks about parallelization

Auto-Detection Algorithm

Step 1: Keyword Analysis

Scan the task description for trigger keywords:

High-Confidence Parallel Triggers:

• "all services" / "every service"
• "all components" / "every component"
• "across all" / "throughout"
• "update all" / "add to every"
• "multiple modules" / "several files"

Sequential Triggers (Do NOT parallelize):

• "migration" / "migrate"
• "depends on" / "after that"
• "then" / "first... then..."
• "in order" / "sequentially"

Codex-Suitable Triggers (Mechanical work):

• "rename" / "find and replace"
• "add docstrings" / "add comments"
• "add logging" / "add caching decorator"
• "boilerplate" / "mechanical"
• "same pattern" / "copy to all"

Step 2: Scope Estimation

Estimate the scope based on task keywords:

Step 3: Dependency Analysis

Check for dependency chains that FORCE sequential execution:

Sequential Chains in UFC-Pokedex:

1. Schema Change Chain: Migration → Load Data → API Update → Type Generation → Frontend
2. Scraper Pipeline: fighters_list → fighter_detail → validation → load
3. Type Safety Chain: Pydantic Schema → FastAPI → OpenAPI → TypeScript Types
4. Test Chain: Unit → Integration → E2E (if E2E depends on data)

Independent Modules (Safe to Parallelize):

• fighter_service ↔ event_service ↔ stats_service (independent)
• FighterCard ↔ EventCard ↔ StatsCard (independent components)
• Backend tests ↔ Frontend tests (independent test suites)
• Scraper spiders for different data types (fighters_list ↔ events_list)

Step 4: Strategy Selection

Based on analysis, select one of four strategies:

Execution Strategies

Strategy A: Parallel Claude Sub-agents

Use superpowers:subagent-driven-development skill

When to use:

• 3+ independent subtasks requiring judgment
• Complex logic or architectural decisions needed
• Code review between tasks is valuable
• Different expertise needed per subtask

UFC-Pokedex examples:

• "Add fighter comparison feature" → 3 agents (API, Service, Frontend)
• "Implement new ranking system" → 2 agents (Backend logic, Frontend display)
• "Add comprehensive tests" → N agents (one per module)

Output format:

I recommend Strategy A: Parallel Claude Sub-agents

Subtasks:
1. [Backend API + Service] - Claude agent with backend context
2. [Frontend Components] - Claude agent with frontend context
3. [Integration Tests] - Claude agent after 1 & 2 complete

Proceed with `superpowers:subagent-driven-development`?

Strategy B: Parallel Codex Agents

Use /​handoffcodex command

When to use:

• Mechanical, repetitive changes across many files
• Clear pattern to apply (no judgment needed)
• Same transformation applied N times
• Bulk operations (rename, add decorator, etc.)

UFC-Pokedex examples:

• "Add @cached decorator to all service methods"
• "Add error handling to all repository methods"
• "Rename fighter_id to fighter_uuid everywhere"
• "Add docstrings to all public functions"

Output format:

I recommend Strategy B: Codex Handoff

This is a mechanical task suitable for Codex:
- Pattern: [describe pattern]
- Files: ~[N] files in [directories]
- Transformation: [describe change]

Proceed with `/​handoffcodex`?

Strategy C: Sequential Implementation

Direct implementation with ordered steps

When to use:

• Clear dependency chain exists
• Later steps depend on earlier results
• Data pipeline execution
• Migration-related changes

UFC-Pokedex examples:

• "Add new field from scraper to API" → Scraper → Migration → Repository → Service → API → Types → Frontend
• "Run weekly scraper and load data" → Scrape → Validate → Load
• "Update database schema" → Migration → Seed → Verify

Output format:

I recommend Strategy C: Sequential Implementation

Dependency chain detected:
1. [Step 1] - must complete first
2. [Step 2] - depends on step 1
3. [Step 3] - depends on step 2
...

I'll implement these in order. Proceed?

Strategy D: Direct Implementation

No parallelization needed

When to use:

• Task touches 1-2 files
• Simple, clear change
• No opportunity for parallelization
• Quick fix or small enhancement

UFC-Pokedex examples:

• "Fix typo in README"
• "Update single component styling"
• "Add one test case"
• "Fix bug in specific function"

Output format:

This is a straightforward task (touches [N] files).
No parallelization needed - I'll implement directly.

UFC-Pokedex Specific Patterns

Backend Patterns

Frontend Patterns

Data Pipeline Patterns

Testing Patterns

MCP Constraint (CRITICAL)

Playwright MCP tasks MUST run sequentially:

• Browser automation cannot be parallelized
• E2E tests must run one at a time
• Screenshots/​snapshots are sequential operations

If task involves Playwright/​MCP:

• Do NOT suggest parallel execution for MCP portions
• Can parallelize non-MCP work alongside sequential MCP work

Decision Output Template

When this skill is invoked, always output:

## Task Analysis

**Task**: [Summarize the task]
**Scope**: [N files / M modules / L layers]
**Pattern Detected**: [Layer-based / Module-based / Pipeline / Simple]
**Dependencies**: [None / Chain: X → Y → Z]

## Recommendation

**Strategy**: [A: Claude Parallel / B: Codex / C: Sequential / D: Direct]
**Reason**: [Why this strategy fits]

### Execution Plan

[If A or B - list subtasks/​agents]
[If C - list ordered steps]
[If D - describe direct approach]

**Proceed with this approach?**

Quick Reference

Task has "all/​every/​multiple" + mechanical change? → Strategy B (Codex)
Task has "all/​every/​multiple" + needs judgment?   → Strategy A (Claude parallel)
Task mentions "migration/​depends/​then/​after"?     → Strategy C (Sequential)
Task touches 1-2 files?                           → Strategy D (Direct)
Task involves Playwright/​MCP?                     → Sequential for MCP portions

Related Resources

• See CLAUDE.md → "Sub-Agent Orchestration" section for project-level guidance
• See superpowers:subagent-driven-development skill for Claude parallel execution
• Use /​handoffcodex command for Codex delegation
• See .claude/​skills/​managing-dev-environment/ for environment setup