Creation Guard

Creation Guard

Prevents duplicate functionality and enforces analysis before creating new Claude Code artifacts.

Purpose

Before creating ANY new:

• Skill (~/.claude/skills/*/SKILL.md)
• Agent (~/.claude/agents/*.md)
• Slash command (~/.claude/commands/*.md)
• CLI tool (~/bin/*)

This skill MUST be invoked to:

1. Search for existing artifacts with similar functionality
2. Identify potential overlap or extension opportunities
3. Present findings with recommendation
4. Get explicit user approval before proceeding

Trigger Phrases

Invoke this skill when you detect:

• "Create a skill for..."
• "I want a new command that..."
• "Let's add an agent for..."
• "Build a CLI tool to..."
• Any intent to create new automation/tooling
• "Could this be a project?"
• "Should we build/make/create..."
• "What if we had a tool that..."
• External feedback identifying a gap (e.g., "do you have a solution for X?")
• Excitement about building something new (warning sign - check ecosystem first)

Analysis Process

Step 0: Check Existing Ecosystem

Before analyzing artifacts, check if this capability already exists elsewhere:

1. Project folders: Look in your project/workbench directories for related work
2. Dev projects: Check ~/Dev/ for existing code that might solve this
3. Capability inventory: Check any capability inventory or system documentation you maintain

# Quick ecosystem check
ls ~/Dev/ | grep -i "[keywords]"
grep -ril "[keywords]" ~/.claude/skills/ ~/.claude/agents/ ~/.claude/commands/ 2>/dev/null | head -10

If existing capability found, STOP. Present the existing solution to the user. Don't proceed to artifact analysis unless they confirm the existing solution doesn't meet their needs.

Step 1: Identify the Proposal

Extract from the request:

• Name: Proposed name for the artifact
• Type: skill | agent | command | cli-tool
• Purpose: What it does (one sentence)
• Key Functions: 3-5 main capabilities
• Keywords: Searchable terms related to functionality

Step 2: Search Existing Artifacts

Run these searches in parallel:

# Skills - search names and descriptions
for skill in ~/.claude/skills/*/SKILL.md; do
  echo "=== $(basename $(dirname $skill)) ==="
  head -20 "$skill"
done

# Agents - search all agent definitions
for agent in ~/.claude/agents/*.md; do
  echo "=== $(basename $agent) ==="
  head -20 "$agent"
done

# Commands - search all command definitions
for cmd in ~/.claude/commands/*.md; do
  echo "=== $(basename $cmd) ==="
  head -10 "$cmd"
done

# CLI tools - list with descriptions
ls -la ~/bin/

Also search by keywords:

grep -ril "[keyword]" ~/.claude/skills/ ~/.claude/agents/ ~/.claude/commands/

Step 3: Research Domain Expertise (Optional)

If you maintain a searchable knowledge base, book index, or reference library, query it for domain-relevant expertise:

# Example: search your knowledge base for relevant resources
# Adapt this to your own search tool or reference system
# my-search-tool "[domain keywords from proposal]" --limit 10

What to look for:

• Existing resources covering the domain (e.g., methodology guides, best practices)
• Frameworks that should inform the approach
• Contrarian views that might suggest alternative designs

How to apply findings:

• Incorporate proven methodologies into skill workflow
• Reference source material in skill documentation
• Use expert terminology and mental models
• Avoid reinventing approaches already well-documented

Document in output:

DOMAIN EXPERTISE FOUND:
- [Source] - [relevant insight]
- [Framework/Methodology] from [Source] - [how to apply]

If no relevant resources found, note "No domain resources found" and proceed.

Step 4: Analyze Overlap

For each potentially related artifact, assess:

Step 5: Generate Recommendation

Based on analysis, recommend ONE of:

Output Format

================================================================
CREATION GUARD ANALYSIS
================================================================

PROPOSAL:
  Type: [skill|agent|command|cli-tool]
  Name: [proposed-name]
  Purpose: [one sentence]

EXISTING ARTIFACTS ANALYZED: [count]

RELATED ARTIFACTS FOUND:

1. [artifact-name] ([type])
   Purpose: [what it does]
   Overlap: [X]% - [explanation]

2. [artifact-name] ([type])
   Purpose: [what it does]
   Overlap: [X]% - [explanation]

DOMAIN EXPERTISE:

- [Source] - [relevant insight or methodology]
- [Source] - [applicable framework or best practice]
(or: No domain resources found)

RECOMMENDATION: [PROCEED|EXTEND|COMPOSE|ITERATE|BLOCK]

RATIONALE:
[2-3 sentences explaining the recommendation]

SUGGESTED ACTION:
[Specific next step based on recommendation]

IF PROCEEDING - INCORPORATE:
[Key insights from domain expertise to build into the artifact]

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

Recommendation Details

PROCEED

• Artifact is genuinely new
• No significant overlap found
• Clear differentiation from existing tools
• Go ahead and create

EXTEND

Present extension proposal:

Instead of creating [new-name], extend [existing-name]:

Current capabilities:
- [existing feature 1]
- [existing feature 2]

Proposed additions:
- [new feature 1]
- [new feature 2]

This approach:
- Maintains single source of truth
- Reduces cognitive load
- Leverages existing testing/documentation

COMPOSE

Present composition approach:

The proposed functionality can be achieved by combining:

1. [artifact-1] - handles [aspect]
2. [artifact-2] - handles [aspect]

Options:
A) Document this workflow (no new code)
B) Create thin orchestration command
C) Add to existing skill's "integration" section

ITERATE

Present refinement questions:

Overlap detected with [existing-artifact].

Differentiation needed:
1. [question about scope]
2. [question about use case]
3. [question about implementation]

Please clarify to proceed.

BLOCK

Explain why creation should not proceed:

BLOCKED: Would create problematic duplication.

Existing artifact: [name]
- Already does: [capabilities]
- Your request: [same capabilities]

Alternative actions:
1. Use existing: /[command] or run [skill]
2. If missing features, extend the existing artifact
3. If different use case, explain how this differs

Integration with Workflow

Hook into Creation Patterns

When Claude detects intent to create new artifacts, it should:

1. Pause - Do not start writing the artifact
2. Invoke - Run this skill's analysis
3. Present - Show findings to user
4. Confirm - Get explicit approval before proceeding
5. Document - If proceeding, note the analysis was done

Self-Check Questions

Before creating ANY new artifact, Claude should ask itself:

1. Does something similar already exist?
2. Could this be added to an existing artifact?
3. Would a user know to look for this vs the existing one?
4. Am I creating this because it's needed or because it's easier than finding what exists?

Examples

Example 1: Duplicate Detection

User: "Create a skill for detecting AI-generated writing patterns"

Analysis finds: slop-detector skill already exists with 95% functional overlap.

Recommendation: BLOCK

• slop-detector already detects AI writing patterns
• Same detection framework, same output format
• User should use existing skill

Example 2: Extension Opportunity

User: "Create a command for logging work to projects"

Analysis finds: log-to-daily skill exists (70% overlap)

Recommendation: EXTEND

• log-to-daily logs to daily notes
• Proposal wants project-specific logging
• Suggest: Add project targeting to existing skill OR create thin wrapper

Example 3: Genuine New Need

User: "Create a skill for managing analytics dashboards"

Analysis finds: No existing analytics skills.

Recommendation: PROCEED

• New capability not covered by existing artifacts
• Clear differentiation from other tools
• Proceed with creation

Success Criteria

This skill is complete when:

• All artifact types searched (skills, agents, commands, cli-tools)
• Related artifacts identified with overlap assessment
• Clear recommendation provided with rationale
• User has explicitly acknowledged the analysis
• If PROCEED: Creation can begin
• If not PROCEED: Alternative action documented