notque/vexjoy-agent / skills/meta/toolkit
toolkit
VerifiedToolkit management: create and evaluate skills and agents, manage routing tables, generate Claude.md.
references/skill-creator/complexity-tiers.md
View on GitHub# Complexity Tier Examples
Real skills from the Claude Code ecosystem categorized by complexity tier with rationale.
## Simple Tier (300-600 lines)
### pr-workflow (cleanup)
**Lines**: ~350
**Purpose**: Local branch cleanup after PR merge
**Phases**: 4 (Identify, Switch, Delete, Prune)
**Why Simple**:
- Single linear workflow
- No subagent coordination
- Minimal error cases (3)
- No reference files needed
**Structure**:
```
.claude/skills/pr-workflow (cleanup)/
└── SKILL.md (350 lines)
- Frontmatter (40 lines)
- Instructions - 4 phases (180 lines)
- Error Handling - 3 errors (60 lines)
- Detection and Fix Patterns - 2 patterns (40 lines)
- Anti-Rationalization (30 lines)
```
---
## Medium Tier (800-1500 lines)
### systematic-debugging
**Lines**: ~1200
**Purpose**: Evidence-based root cause analysis
**Phases**: 4 (Reproduce, Isolate, Identify, Verify)
**Why Medium**:
- Multi-phase workflow with gates
- Moderate error handling (7 cases)
- Some failure modes (4)
- No references/ needed yet
**Structure**:
```
.claude/skills/systematic-debugging/
└── SKILL.md (1200 lines)
- Frontmatter (60 lines)
- Instructions - 4 phases (500 lines)
- Error Handling - 7 errors (250 lines)
- Detection and Fix Patterns - 4 patterns (200 lines)
- Anti-Rationalization (100 lines)
- Blocker Criteria (90 lines)
```
---
### pr-workflow (commit intent)
**Lines**: ~1100
**Purpose**: Phase-gated git commit workflow (now the `commit` intent of the pr-workflow umbrella)
**Phases**: 5 (Status, Diff, Log, Stage, Commit)
**Why Medium**:
- Sequential Git operations
- Validation at each phase
- Moderate scripting (2 scripts)
- 6 error cases
**Structure**:
```
.claude/skills/process/pr-workflow/
├── SKILL.md
├── references/commit.md (~900 lines)
└── scripts/
├── validate_state.py
└── validate_message.py
```
---
## Complex Tier (1500-2500 lines)
### parallel-code-review
**Lines**: ~2200
**Purpose**: Parallel 3-reviewer orchestration
**Phases**: 4 (Prepare, Execute, Aggregate, Report)
**Why Complex**:
- Multi-agent coordination (3 reviewers)
- Parallel execution with timeouts
- Verdict synthesis logic
- Death loop prevention required
- 10+ error cases
**Structure**:
```
.claude/skills/review/parallel-code-review/
├── SKILL.md (1600 lines)
│ - Frontmatter (80 lines)
│ - Instructions - 4 phases (700 lines)
│ - Death Loop Prevention (200 lines)
│ - Error Handling - top 5 (150 lines)
│ - Detection and Fix Patterns - top 5 (200 lines)
│ - Anti-Rationalization (150 lines)
│ - References (120 lines)
└── references/
├── error-catalog.md (400 lines)
└── verdict-aggregation.md (200 lines)
```
---
### workflow-orchestrator
**Lines**: ~2100
**Purpose**: Three-phase task orchestration
**Phases**: 4 (BRAINSTORM, WRITE-PLAN, EXECUTE-PLAN, VERIFY)
**Why Complex**:
- Multi-file coordination
- Task tool integration
- State management
- Multiple workflow patterns
- Extensive error handling
**Structure**:
```
.claude/skills/workflow/references/workflow-orchestrator.md
├── SKILL.md (1500 lines)
│ - Frontmatter (70 lines)
│ - Instructions - 4 phases (800 lines)
│ - State Management (200 lines)
│ - Error Handling - top 5 (150 lines)
│ - Detection and Fix Patterns - top 5 (180 lines)
│ - References (100 lines)
└── references/
├── error-catalog.md (350 lines)
└── workflow-patterns.md (250 lines)
```
---
## Comprehensive Tier (2500-4000 lines)
### go-patterns
**Lines**: ~3800
**Purpose**: Go testing patterns and methodology
**Phases**: Multiple workflows (table-driven, subtests, helpers, mocks, benchmarks)
**Why Comprehensive**:
- Multiple complex workflows
- Extensive code examples (30+)
- Deep Go testing expertise
- Comprehensive error catalog
- Reference-quality documentation
**Structure**:
```
.claude/skills/engineering/go-patterns/
├── SKILL.md (2000 lines)
│ - Frontmatter (90 lines)
│ - Core Workflows (800 lines)
│ - Top 5 errors (200 lines)
│ - Top 5 anti-patterns (250 lines)
│ - Quick reference (300 lines)
│ - References section (160 lines)
└── references/
├── error-catalog.md (600 lines)
├── code-examples.md (800 lines)
├── preferred-patterns.md (500 lines)
└── table-driven-testing.md (400 lines)
```
---
### go-patterns
**Lines**: ~3500
**Purpose**: Go concurrency patterns and primitives
**Phases**: Multiple workflows (goroutines, channels, sync primitives, worker pools)
**Why Comprehensive**:
- Multiple complex concurrency patterns
- 40+ code examples
- Extensive race condition catalog
- Deep Go concurrency expertise
- Production debugging patterns
**Structure**:
```
.claude/skills/engineering/go-patterns/
├── SKILL.md (1800 lines)
│ - Frontmatter (100 lines)
│ - Core Patterns (700 lines)
│ - Top 5 errors (180 lines)
│ - Top 5 anti-patterns (220 lines)
│ - Quick reference (350 lines)
│ - References section (150 lines)
└── references/
├── error-catalog.md (700 lines)
├── code-examples.md (900 lines)
├── preferred-patterns.md (600 lines)
└── worker-pools.md (400 lines)
```
---
## Tier Selection Decision Tree
### Start: What kind of workflow?
**Single, focused operation?**
→ Simple tier
- Examples: pr-workflow (cleanup), service-health-check, github-notification-triage
- Characteristics: Linear workflow, minimal scripting, <5 errors
**Multi-step with moderate coordination?**
→ Medium tier
- Examples: systematic-debugging, pr-workflow (commit), pr-workflow (fix)
- Characteristics: 2-4 phases, moderate scripting, 5-10 errors
**Multi-agent coordination OR parallel execution?**
→ Complex tier
- Examples: parallel-code-review, workflow-orchestrator, research-coordinator
- Characteristics: Subagent spawning, parallel tasks, death loop prevention, 10+ errors
**Reference-quality with multiple complex workflows?**
→ Comprehensive tier
- Examples: go-patterns, go-patterns, go-patterns
- Characteristics: Multiple workflows, 30+ code examples, extensive catalogs
---
## Migration Patterns
### Simple → Medium (when to upgrade)
**Signals**:
- Added 2nd or 3rd phase with complex gates
- Error cases approaching 10
- Users requesting more detailed workflows
**Example**: pr-workflow (cleanup) → pr-workflow (pipeline) (added commit, push, create-pr phases)
---
### Medium → Complex (when to upgrade)
**Signals**:
- Added subagent coordination
- Parallel execution needed
- Reference files created
- Main file approaching 1500 lines
**Example**: code-review → parallel-code-review (added 3 parallel reviewers)
---
### Complex → Comprehensive (when to upgrade)
**Signals**:
- Multiple distinct workflows in one skill
- Code examples exceeding 20
- Reference files exceeding 1500 total lines
- Becoming reference documentation
**Example**: testing-patterns → go-patterns (added table-driven, mocks, benchmarks workflows)
---
## Complexity Tier Selection Guidelines
### Match Tier to Actual Complexity
**Signal**: Simple workflow (pr-workflow (cleanup)) implemented as Complex tier with references/, death loop prevention, 10+ error cases
**Why this matters**: Creates maintenance burden, confuses users, violates Over-Engineering Prevention
**Preferred action**: Reduce to appropriate tier - Simple workflows stay Simple
---
### Add Required Patterns for Complex Skills
**Signal**: Multi-agent orchestration skill (parallel-code-review) implemented as Medium tier with no death loop prevention
**Why this matters**: Missing critical patterns, will fail in production, no scaling considerations
**Preferred action**: Upgrade to Complex tier, add death loop prevention, create references/
---
### Restructure When Line Count Exceeds Tier Budget
**Signal**: Skill starts as Simple (400 lines), grows to 2000 lines over time without restructuring
**Why this matters**: Violates progressive disclosure, bloats context, makes skill hard to maintain
**Preferred action**: Promote to appropriate tier, extract references/, apply progressive disclosure
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |