Claude Creation Guard

A Claude Code skill that prevents duplicate functionality by analyzing existing artifacts before creating new ones.

The Problem

When working with Claude Code over time, you accumulate skills, agents, commands, and CLI tools. Without guardrails, it's easy to:

• Create a new skill that duplicates an existing one
• Build commands with overlapping functionality
• End up with multiple ways to do the same thing
• Forget what tools you already have

This skill enforces "look before you leap" — analyzing your existing artifacts before creating new ones.

How It Works

When you ask Claude to create a new skill, agent, command, or CLI tool, this skill:

1. Searches all existing artifacts in ~/.claude/
2. Analyzes overlap with the proposed functionality
3. Recommends one of five actions:
   • PROCEED — Genuinely new, go ahead
   • EXTEND — Add to existing artifact instead
   • COMPOSE — Combine existing tools
   • ITERATE — Refine the proposal
   • BLOCK — Would create problematic duplication
4. Waits for explicit user approval before creating anything

Installation

Option 1: Skills CLI

npx skills add aplaceforallmystuff/claude-creation-guard

Option 2: Copy to your Claude Code skills directory

# Clone the repository
git clone https://github.com/aplaceforallmystuff/claude-creation-guard.git

# Copy to your Claude Code skills directory
cp -r claude-creation-guard/skills/creation-guard ~/.claude/skills/

Option 3: Clone directly to skills directory

git clone https://github.com/aplaceforallmystuff/claude-creation-guard.git ~/.claude/skills/creation-guard

Option 4: Manual installation

1. Create the directory: mkdir -p ~/.claude/skills/creation-guard
2. Download SKILL.md to that directory

Usage

The skill activates when Claude detects intent to create new artifacts:

• "Create a skill for..."
• "I want a new command that..."
• "Let's add an agent for..."
• "Build a CLI tool to..."

Example Output

════════════════════════════════════════════════════════════════
CREATION GUARD ANALYSIS
════════════════════════════════════════════════════════════════

PROPOSAL:
  Type: skill
  Name: ai-detector
  Purpose: Detect AI-generated content in text

EXISTING ARTIFACTS ANALYZED: 45

RELATED ARTIFACTS FOUND:

1. slop-detector (skill)
   Purpose: Detect AI-generated writing patterns
   Overlap: 95% - Same detection framework

RECOMMENDATION: BLOCK

RATIONALE:
slop-detector already provides comprehensive AI writing detection
with tiered patterns and scoring. Creating ai-detector would
duplicate this functionality.

SUGGESTED ACTION:
Use the existing slop-detector skill, or extend it if specific
features are missing.

════════════════════════════════════════════════════════════════
Proceed with creation? (y/n/discuss)
════════════════════════════════════════════════════════════════

Recommendation Types

Recommendation	When Used	Action
PROCEED	<20% overlap, genuinely new	Create the artifact
EXTEND	50%+ overlap with one artifact	Modify existing instead
COMPOSE	Multiple artifacts cover 80%+	Create thin wrapper or document workflow
ITERATE	20-50% overlap	Refine proposal to differentiate
BLOCK	Would duplicate existing	Do not create

Integration with CLAUDE.md

For automatic enforcement, add this to your CLAUDE.md:

## Artifact Creation Guard (MANDATORY)

Before creating ANY new skill, agent, command, or CLI tool, MUST invoke the `creation-guard` skill.

**Trigger phrases that require this check:**
- "Create a skill for..."
- "Let's add a command that..."
- "Build an agent for..."
- "Make a CLI tool to..."

**Process:**
1. STOP before writing any new artifact
2. Run creation-guard analysis
3. Search existing skills, agents, commands for overlap
4. Present findings with recommendation
5. Get explicit user approval
6. Only then proceed (or take alternative action)

Related Skills

Part of the aplaceforallmystuff skills collection:

• ecosystem-health — Detect drift across your entire Claude Code setup
• think-first — Mental model application before significant decisions
• rfu-audit — 11-gate utility validation before investing effort

License

MIT License - see LICENSE

---

The goal: One tool per job, clearly named, easy to find.