---

name: your-skill-name description: > Concise description of what this skill does (the "what") and when to use it (the "when"). Include trigger keywords naturally. Max 1024 chars, third person voice. Example: "Processes Excel files to generate monthly reports with charts. Use when analyzing sales data, creating spreadsheets, or generating reports." version: "1.0.0" license: Apache-2.0 user-invocable: true # Show in slash menu? (default: true)

disable-model-invocation: false # Prevent auto-trigger? Uncomment for manual-only skills

mode: interactive # Uncomment if skill requires user interaction

allowed-tools:

• Read
• Write
• Bash

Add other tools as needed: Edit, Glob, Grep, Task, etc.

triggers:

Keywords (1-2 words) - Core concepts users mention

• "keyword"
• "mot-clé"
• "concept"

Action phrases (FR + EN) - What users want to DO

• "créer quelque chose"
• "create something"
• "faire une action"
• "do an action"

Problem phrases - How users describe their PROBLEM

• "j'ai besoin de..."
• "I need to..."
• "comment faire pour..."
• "how do I..."

Add 10-20 natural language triggers total

Include both French and English variants

Use phrases users naturally speak, not technical jargon

metadata: author: "Your Name" category: "category-name" keywords: ["keyword1", "keyword2"]

Your Skill Name

Brief overview of what this skill does (1-2 sentences).

When to Use

Describe specific scenarios where this skill should be invoked:

• Scenario 1: When user asks about X
• Scenario 2: When user needs to Y
• Scenario 3: When working on Z

Observability

REQUIRED: Every skill MUST announce its activation at the start for observability.

First: At the beginning of execution, display:

Copy🔧 Skill "your-skill-name" activated

This confirms which skill is running and provides feedback to users.

Prerequisites

List any requirements before using this skill:

• Installed software (e.g., Python 3.8+, Node.js)
• Environment variables needed
• Access permissions required
• Other skills that should run first

Execution Steps

Detail the step-by-step process Claude should follow:

Step 1: Preparation

Describe what to prepare or validate first.

Copy# Example command if needed
echo "Preparation step"

Step 2: Main Processing

Describe the core functionality.

Important considerations:

• Edge case 1: How to handle X
• Edge case 2: What to do if Y fails
• Edge case 3: When Z condition occurs

Step 3: Validation

Verify results and ensure correctness.

Copy# Validation command example
test -f output.txt && echo "Success" || echo "Failed"

Step 4: Completion

Final steps, cleanup, or reporting.

Expected Output

Describe the format and content of the output:

Format: File type, structure, or response format

Content:

• What the output contains
• How it's organized
• Where it's saved

Example output:

CopyExample of what the output looks like

Error Handling

Common errors and how to resolve them:

Troubleshooting steps:

1. Check prerequisites are met
2. Verify environment variables are set
3. Review logs in [location]
4. Try running with verbose mode: command --verbose

Examples

Example 1: Basic Usage

User request: "Do something simple"

Steps:

1. Execute basic command
2. Verify result
3. Report status

Result:

CopyExpected output for basic usage

Example 2: Advanced Usage

User request: "Do something complex with options"

Steps:

1. Parse options
2. Execute with parameters
3. Handle edge cases
4. Validate complex result

Result:

CopyExpected output for advanced usage

Skill Chaining

Skills Required Before

• skill-name-1: Why it's needed first
• skill-name-2: What it provides
• Or: "None (entry point skill)"

Input Expected

• Format: Description of input format
• Source: Where input comes from (user, file, previous skill)
• Validation: How to verify input is valid

Output Produced

• Format: Description of output format (file, response, side effect)
• Location: Where output is saved or sent
• Side effects: Any state changes, files created, services modified
• Duration: Expected time to complete

Compatible Skills After

Recommended:

• skill-name-1: What it does with this output
• skill-name-2: How it extends this workflow

Optional:

• skill-name-3: Alternative next step

Called By

• skill-name-1: In what context
• Or: "Direct user invocation"
• Or: "Background scheduler"

Tools Used

• Read: Read configuration files, input data
• Write: Create output files, reports
• Bash: Execute scripts, run commands
• Edit: Modify existing files
• (List all tools used with brief explanation)

Visual Workflow

CopyUser Request / Previous Skill Output
    ↓
[THIS SKILL]
    ├─► Step 1: Preparation
    ├─► Step 2: Main Processing
    ├─► Step 3: Validation
    └─► Step 4: Completion
    ↓
Output File / Next Skill Input
    ↓
[Optional] Compatible Skills After

Usage Example

Scenario: Real-world use case description

Input: What's provided

CopyExample input data or command

Command: How to invoke (if applicable)

Copy# Example invocation
your-command --options input.txt

Output: What's produced

CopyExample output or result

Next steps: What to do with the output

References

If you need to reference additional documentation, create files in references/ directory:

• references/detailed-guide.md - Detailed documentation
• references/api-reference.md - API specifications
• references/examples.md - Additional examples

Files > 100 lines MUST have table of contents at top.

Scripts

If you need executable code, create files in scripts/ directory:

• scripts/process.py - Main processing script
• scripts/validate.sh - Validation script
• scripts/deploy.js - Deployment script

Scripts are NEVER loaded into context - they are executed only when needed.

Assets

If you need templates or output resources, create files in assets/ directory:

• assets/template.xlsx - Excel template
• assets/config-sample.json - Configuration template
• assets/diagram.png - Workflow diagram

Assets are used in output generation, not loaded into context.

---

Template Usage Instructions

To use this template:

1. Copy this folder and rename with your skill name (kebab-case)

2. Modify YAML frontmatter:

   • Set name (lowercase, hyphens, max 64 chars)
   • Write clear description (what + when, max 1024 chars)
   • Add 10-20 triggers (natural language, bilingual FR/EN)
   • Set version (semantic versioning)
   • List allowed-tools (only what's needed)
   • Update metadata (author, category, keywords)

3. Replace placeholder content:

   • Update all "your-skill-name" references
   • Write When to Use scenarios
   • Detail Execution Steps
   • Document Expected Output
   • Add Error Handling table
   • Provide concrete Examples
   • Complete Skill Chaining section

4. Keep SKILL.md < 500 lines:

   • Move detailed docs to references/
   • Move code to scripts/
   • Move templates to assets/

5. Test triggers:

   Copypython scripts/benchmark-semantic-router.py "your test phrase"
   

6. Review quality: Use julien-skill-reviewer to score and improve your skill.

Delete this section after creating your skill!