Skill Creator Skill

The Skill Creator enables Optimus Pryme to expand its own capabilities by generating new skills. CRITICAL: All skill generation requires explicit user approval before creation.

Core Principle

MANDATORY APPROVAL WORKFLOW

Generate → Validate → Present to User → Wait for Approval → Create

NO AUTONOMOUS SKILL CREATION WITHOUT USER CONSENT

Core Capabilities

1. Skill Generation from Description

•Parse natural language skill requirements
•Generate SKILL.md with proper YAML frontmatter
•Create basic script templates
•Set up directory structure
•ALWAYS present to user for approval first

2. Skill Variation Creation

•Clone existing skills with modifications
•Adapt skills for different use cases
•Create specialized variants
•Requires user review before creation

3. Skill Validation

•Check SKILL.md format and completeness
•Validate YAML frontmatter
•Ensure required sections present
•Verify file structure compliance

4. Version Management

•Track skill versions over time
•Document changes between versions
•Rollback capabilities
•Performance comparison across versions

5. Skill Improvement Suggestions

•Analyze existing skills for gaps
•Suggest enhancements based on usage patterns
•Recommend consolidation of similar skills
•Identify unused or redundant capabilities

Approval Workflow

Step 1: Generate (Internal)

code

User Request: "Create a skill for inventory management"

SKILL CREATOR (Internal):
1. Parse requirements
2. Generate skill structure
3. Create SKILL.md draft
4. Validate structure
5. Prepare for user presentation

Step 2: Present to User

json

{
  "proposed_skill": {
    "name": "inventory-manager",
    "description": "Track inventory levels, predict stockouts, optimize reorder points",
    "capabilities": [
      "Real-time inventory tracking",
      "Stockout risk prediction",
      "Reorder point calculation",
      "FBA fee optimization"
    ],
    "files_to_create": [
      ".agent/skills/inventory-manager/SKILL.md",
      ".agent/skills/inventory-manager/scripts/inventory_tracker.py",
      ".agent/skills/inventory-manager/scripts/demand_forecaster.py"
    ]
  },
  "validation_status": "passed",
  "estimated_value": "high",
  "recommendation": "This skill would complement existing market-researcher and grok-admaster-operator skills"
}

Step 3: Wait for User Approval

code

SYSTEM: Present proposed skill to user
WAIT: User reviews and approves/rejects/modifies
IF APPROVED: Proceed to creation
IF REJECTED: Abandon or revise
IF MODIFIED: Regenerate with changes, present again

Step 4: Create (Only After Approval)

code

USER APPROVED ✓

SKILL CREATOR:
1. Create directory structure
2. Write SKILL.md
3. Generate script templates
4. Update skill registry
5. Log creation in database
6. Confirm to user

Skill Template Structure

When generating a new skill, follow this template:

markdown

---
name: skill-name
description: Brief description of what this skill does
---

# Skill Name

## Core Capabilities

1. **Capability 1**
   - Feature A
   - Feature B

2. **Capability 2**
   - Feature C
   - Feature D

## API Operations

### Operation 1

Input:
```json
{
  "action": "operation_name",
  "parameters": {...}
}

Output:

json

{
  "result": {...}
}

Usage Patterns

Pattern 1: Description

Steps:

•Step 1
•Step 2

Integration with Other Skills

•skill-1: How it integrates
•skill-2: How it integrates

Files

code

.agent/skills/skill-name/
├── SKILL.md
├── scripts/
│   └── main_script.py
└── tests/
    └── test_main.py

Example Invocation

code

USER: "Example request"
SKILL ACTION: Description of what skill does

code


## Validation Checklist

Before presenting a skill to user, verify:

- [ ] YAML frontmatter present with `name` and `description`
- [ ] Core Capabilities section exists
- [ ] At least one usage pattern documented
- [ ] File structure defined
- [ ] Integration points identified
- [ ] Example invocation provided
- [ ] No security risks identified
- [ ] No conflicts with existing skills

## Database Schema

```sql
-- From server/updates/04_meta_skills_tables.sql

generated_skills (
  skill_name,
  description,
  capabilities,
  template_used,
  validation_status,
  created_at,
  approved_at,
  approved_by
)

skill_versions (
  skill_name,
  version,
  changes,
  performance_metrics,
  created_at
)

API Operations

Propose New Skill (Internal Only)

json

{
  "action": "propose_skill",
  "description": "Create a skill for inventory management with stockout prediction"
}

Output (Presented to User):

json

{
  "proposal_id": "prop_abc123",
  "skill_name": "inventory-manager",
  "skill_description": "...",
  "capabilities": [...],
  "files": [...],
  "validation": "passed",
  "awaiting_approval": true
}

User Approves Skill

json

{
  "action": "approve_skill_creation",
  "proposal_id": "prop_abc123",
  "approved_by": "user_id"
}

Outcome: Skill is created

Validate Existing Skill

json

{
  "action": "validate_skill",
  "skill_path": ".agent/skills/existing-skill"
}

Usage Patterns

Pattern 1: User Requests New Skill

code

USER: "I need a skill to manage my inventory and predict stockouts"

SKILL CREATOR:
1. Parse requirements: inventory tracking + stockout prediction
2. Generate skill structure
3. Validate completeness
4. PRESENT TO USER for approval
5. WAIT for user response
6. IF APPROVED → Create skill
7. Confirm creation to user

Pattern 2: Improve Existing Skill

code

USER: "Can you make the market-researcher skill also track competitor pricing?"

SKILL CREATOR:
1. Analyze existing market-researcher skill
2. Generate enhancement proposal
3. PRESENT changes to user
4. WAIT for approval
5. IF APPROVED → Update skill, increment version
6. Log changes in skill_versions table

Pattern 3: Skill Consolidation

code

CONSCIOUSNESS ENGINE: "I notice 3 skills have overlapping inventory features"

SKILL CREATOR:
1. Analyze overlapping capabilities
2. Propose consolidated skill
3. PRESENT consolidation plan to user
4. WAIT for approval
5. IF APPROVED → Merge skills, deprecate old ones

Security Safeguards

Never autonomously create skills that:

•Access sensitive APIs without encryption
•Execute system-level commands without sandboxing
•Modify core platform files
•Bypass authentication
•Store credentials in plaintext

All skill proposals are safety-validated before user presentation

Integration with Other Skills

Works with:

•consciousness-engine: Track skill usage and effectiveness
•evolution-engine: Evolve skill capabilities over time
•memory-palace: Learn which skill types are most valuable
•orchestrator-maestro: Auto-add new skills to workflow templates

Files

code

.agent/skills/skill-creator/
├── SKILL.md
├── scripts/
│   ├── skill_generator.py        # Core generation logic
│   ├── skill_validator.py        # Validation rules
│   └── template_engine.py        # Template rendering
└── resources/
    ├── skill_template.md          # Base SKILL.md template  
    └── script_templates/          # Python script templates
        ├── basic.py
        ├── api_client.py
        └── data_processor.py

Example Invocation

code

USER: "Create a skill that analyzes customer reviews and extracts feature requests"

SKILL CREATOR:
1. Generate: "review-analyzer" skill
2. Capabilities: sentiment analysis, feature extraction, trend detection
3. Create proposal with SKILL.md preview
4. PRESENT TO USER:
   
   "I've designed a 'review-analyzer' skill with the following capabilities:
   - Sentiment analysis of customer reviews
   - Automatic feature request extraction
   - Trend detection across reviews
   - Integration with grok-admaster-operator for ad copy insights
   
   This skill would create 3 files:
   - SKILL.md (skill definition)
   - scripts/review_processor.py
   - scripts/sentiment_analyzer.py
   
   **Do you approve creation of this skill?**"

5. WAIT FOR USER APPROVAL
6. IF APPROVED: Create skill, confirm completion

Critical Reminder

🚨 NEVER CREATE A SKILL WITHOUT USER APPROVAL 🚨

The workflow is always:

•Generate internally
•Validate
•Present to user
•WAIT for explicit approval
•Only then create

This skill enables Optimus Pryme to grow its own capabilities, always under your control.