skill

name	skill
description	Create new skill from documentation, llms.txt, or GitHub repo

🎯 SKILL CREATOR

Create new skill:

Input: {{args}}

INPUT TYPES

Type	Example
Natural language	/skill create MongoDB integration skill
llms.txt URL	/skill create https://docs.example.com/llms.txt
GitHub repo	/skill create https://github.com/org/repo

---

4-PHASE WORKFLOW

Phase 1: Research (2-3 min)

• Fetch documentation (llms.txt, GitHub, web)
• Analyze domain knowledge required
• Identify key concepts and patterns
• Map out skill structure
• Research best practices

Phase 2: Implementation (3-5 min)

• Write skill prompt file
• Structure knowledge sections
• Create usage examples
• Define when to use skill
• Include best practices

Phase 3: Review (1-2 min)

• Validate skill structure
• Check completeness
• Verify examples work
• Ensure clear instructions

Phase 4: Evaluation (1 min)

• Test skill activation
• Verify skill prompt loads
• Check examples are clear

---

SKILL FILE STRUCTURE

Create skill file at: ~/.gemini-kit/skills/[skill-name].md

# [Skill Name] Skill

## When to Use
- [Condition 1]
- [Condition 2]

## Key Concepts

### [Topic 1]
[Knowledge content]

### [Topic 2]
[Knowledge content]

## Examples

### Example 1: [Scenario]
[Step-by-step example]

### Example 2: [Scenario]
[Step-by-step example]

## Best Practices
- [Practice 1]
- [Practice 2]

## Common Patterns
[Code patterns and templates]

## Troubleshooting
| Issue | Solution |
|-------|----------|
| [Issue 1] | [Solution 1] |

---

COMMON SKILL TYPES

Type	Examples
Framework	React, Vue, Next.js, Flutter
Cloud	AWS, GCP, Azure, Vercel
Database	MongoDB, PostgreSQL, Redis
Tools	Docker, Kubernetes, Terraform
Domain	Auth, Payment, Email, AI/ML

---

OUTPUT FILES

After creation:

1. Skill file: ~/.gemini-kit/skills/[skill-name].md
2. Activation: Skill auto-activates on matching keywords

---

USAGE EXAMPLES

# From description
/skill create MongoDB integration with Node.js

# From llms.txt
/skill create https://docs.stripe.com/llms.txt

# From GitHub
/skill create https://github.com/prisma/prisma

---

FIX-LOGS MODE

When input starts with "fix-logs" or "fix":

Syntax:

/skill fix-logs [skill-name]
/skill fix mongodb-skill

5-Step Workflow:

Step 1: Log Analysis

• Read logs.txt file
• Identify skill-related errors
• Extract error messages and context
• Map errors to skill sections

Step 2: Skill Diagnosis

• Analyze current skill content
• Compare with error patterns
• Identify root causes
• Find incorrect examples

Step 3: Fix Planning

• List issues to address
• Plan content additions
• Identify sections to rewrite

Step 4: Skill Update

• Update problematic sections
• Add missing information
• Fix code examples
• Improve error handling guidance

Step 5: Validation

• Verify fixes address logged issues
• Test updated examples
• Validate skill structure

Output:

✓ Analyzed: 15 errors in logs.txt
✓ Identified: 3 skill issues
  - Incorrect webhook verification code
  - Missing SDK import example
  - Outdated API endpoint

✓ Fixed: mongodb-skill.md
  - Updated webhook example
  - Added SDK import section
  - Fixed API endpoints

Validation: PASSED

---

ADD MODE

When input starts with "add":

Syntax:

/skill add [skill-name] [reference]
/skill add mongodb https://docs.mongodb.com/llms.txt
/skill add stripe /path/to/examples.md

Reference Types:

Type	Example
URL	https://docs.stripe.com/api
File	/path/to/reference.md
Script	validate-webhook.sh

5-Step Workflow:

Step 1: Validate Skill

• Check skill exists
• Load current skill structure

Step 2: Analyze Reference Type

• Detect URL/File/Script
• Validate format

Step 3: Process Reference

• URLs: Fetch and extract content
• Files: Read and validate
• Scripts: Create executable

Step 4: Update Skill

• Add reference to skill file
• Structure for token efficiency
• Core content vs on-demand

Step 5: Test Activation

• Verify skill still loads
• Test new references accessible

Output:

Adding reference to mongodb skill...
✓ Skill found: ~/.gemini-kit/skills/mongodb.md
✓ Reference: https://docs.mongodb.com/llms.txt
✓ Content: 12KB, 45 endpoints
✓ Updated: mongodb.md

New sections added:
- API Endpoints (45)
- Configuration Options (12)
- Examples (8)

---

OPTIMIZE MODE

When input starts with "optimize":

Syntax:

/skill optimize [skill-name] "[goal]"
/skill optimize mongodb "reduce token usage by 40%"
/skill optimize stripe "improve clarity"

Optimization Goals:

Goal	Action
Token reduction	Remove duplicates, compress
Clarity	Better instructions
Examples	Add/improve examples
Progressive disclosure	On-demand loading

7-Step Workflow:

Step 1: Read Skill Files

• Load skill content
• Count tokens

Step 2: Analyze Against Goal

• Find optimization opportunities
• Identify duplicates, verbose text

Step 3: Create Plan

• Token reduction targets
• Risk assessment

Step 4: Present to User

OPTIMIZATION PLAN
Skill: mongodb
Goal: Reduce tokens by 40%

Current: 8,400 tokens
Target:  5,040 tokens

Proposed Changes:
1. Consolidate docs: -1,200 tokens
2. Progressive disclosure: -1,400 tokens
3. Compress instructions: -800 tokens

Estimated: 5,000 tokens (40.5% reduction)

Step 5: Wait for Approval

• User confirms changes

Step 6: Execute

• Apply optimizations
• Update skill file

Step 7: Test Optimized Skill

• Verify skill still works
• Count new token usage

Output:

✓ Optimized: mongodb skill
  Before: 8,400 tokens
  After:  5,000 tokens
  Saved:  3,400 tokens (40.5%)

---

OPTIMIZE:AUTO MODE

When input starts with "optimize:auto" or "optimize-auto":

Syntax:

/skill optimize:auto [skill-name] "[goal]"
/skill optimize:auto mongodb "reduce tokens"

Difference from /skill optimize:

Feature	optimize	optimize:auto
Approval	✅ Required	❌ No
Backup	❌	✅ Auto
Rollback	❌	✅ Auto

6-Step Workflow:

Step 1: Create Backup

Backup: ~/.gemini-kit/skills/mongodb/.backup-251215-2110/

Step 2: Analyze Goal

• Find optimization opportunities

Step 3: Apply Changes

• No approval needed
• Apply immediately

Step 4: Validate Syntax

• Check skill still valid

Step 5: Test Activation

• Verify skill loads

Step 6: Report or Rollback

• Success: Report changes
• Failure: Auto rollback from backup

Output:

✓ Backup created
✓ Analyzed: 3 opportunities found
✓ Applied: 3 changes
  - Consolidated docs: -800 tokens
  - Progressive disclosure: -1,200 tokens
  - Compressed instructions: -400 tokens
✓ Syntax valid
✓ Activation successful

Result: 6,400 → 4,000 tokens (37.5% saved)

Rollback (if error):

❌ Activation failed
✓ Rolling back from backup...
✓ Restored original skill

---

Now process: "{{args}}"