Skill Requirements Gathering Workflow

This workflow defines how to gather requirements for creating a new skill.

Required Information

1. Skill Name - Identifier (kebab-case)
2. Description - Brief purpose statement
3. Responsibility - Primary responsibility (1-2 sentences)
4. Tools - Claude Code tools to use
5. Plugin Name - Target plugin location
6. Inputs - What the skill receives
7. Outputs - What the skill returns

Optional Information

8. Handler Type - If multi-provider (e.g., "iac", "hosting")
9. Workflow Steps - High-level steps skill will execute
10. Scripts Needed - List of scripts for deterministic operations
11. Completion Criteria - Success conditions
12. Invokes - Other skills or commands this skill calls

Gathering Process

1. Check Provided Parameters

Check which parameters were already provided in provided_params:

• If skill_name is provided: Use it (already validated by command)
• If description is provided: Use it
• If tools is provided: Use it, otherwise default to "Bash"
• If plugin_name is provided: Validate plugin exists
• If handler_type is provided: Note multi-provider skill

2. Prompt for Missing Required Information

Description (if not provided)

Please provide a brief description of the skill:

This description will appear in the frontmatter and documentation.

Examples:
  • "Fetches data from external sources and validates format"
  • "Deploys infrastructure using Terraform"
  • "Validates code quality and runs linters"

Enter description:

Validation:

• Not empty
• Between 10-200 characters
• Starts with capital letter
• Clear and specific

Responsibility (always prompt)

What is the primary responsibility of this skill?

This is a 1-2 sentence statement of what the skill does.

Examples:
  • "fetching data from external APIs and validating response format"
  • "executing Terraform operations to provision infrastructure"
  • "running code quality checks and reporting violations"

Enter responsibility:

Validation:

• Not empty
• Starts with lowercase (will be inserted after "responsible for")
• Ends without period (formatting handled by template)

Inputs (always prompt)

What inputs does this skill receive?

Describe the parameters passed to the skill when invoked.

Examples:
  • "source URL, authentication credentials, timeout value"
  • "infrastructure specification, environment name, provider config"
  • "file paths to check, linter configuration"

Enter inputs (or "none" if skill takes no parameters):

Validation:

• Not empty
• Clear parameter list or "none"

Outputs (always prompt)

What does this skill return as output?

Describe the structure of the response returned by the skill.

Examples:
  • "JSON object with fetched data and validation status"
  • "Deployment status, resource IDs, and operation logs"
  • "Quality score, list of violations, and recommendations"

Enter outputs:

Validation:

• Not empty
• Clear output description

3. Prompt for Plugin Name (if not provided)

Which plugin should this skill be added to?

Detected plugins in this repository:
  • faber
  • faber-app
  • faber-cloud
  • work
  • repo
  [list all plugins found in plugins/ directory]

Enter plugin name (or press Enter to detect from current directory):

Validation:

• Plugin directory must exist in plugins/
• If empty, detect from current working directory
• Skill will be created in plugins/{plugin_name}/skills/{skill_name}/

4. Prompt for Tools (if not provided)

Which Claude Code tools should this skill use?

Common combinations:
  • "Bash" - Standard for most skills (run scripts)
  • "Bash, Read, Write" - For file manipulation skills
  • "Bash, WebFetch" - For API/web interaction skills
  • "Bash, Grep, Glob" - For file search skills

Enter tools (comma-separated) or press Enter for default [Bash]:

Validation:

• If empty, use default: "Bash"
• Valid tools: Bash, Read, Write, Edit, WebFetch, Grep, Glob, Skill
• Comma-separated list
• Trim whitespace

5. Prompt for Optional Information

Handler Type (if not already provided)

Is this a multi-provider handler skill? (Optional)

Handler skills provide abstraction across multiple providers.
Examples: iac (terraform/pulumi/cdk), hosting (aws/gcp/azure)

Enter handler type or press Enter to skip:

Validation:

• If provided, should be lowercase single word
• Common types: iac, hosting, storage, messaging, monitoring

Workflow Steps (optional)

What are the high-level steps this skill will execute? (Optional)

List the major steps in order.

Examples:
  • "1. Validate input, 2. Fetch data, 3. Transform, 4. Return result"
  • "1. Load config, 2. Plan changes, 3. Apply, 4. Verify"

Enter steps or press Enter to skip:

Scripts Needed (optional)

What scripts does this skill need? (Optional)

List script names for deterministic operations.

Examples:
  • "fetch-data.sh, validate-format.sh"
  • "terraform-plan.sh, terraform-apply.sh"

Enter script names (comma-separated) or press Enter to skip:

Completion Criteria (optional but recommended)

What are the success conditions for this skill? (Optional)

Define what "complete and successful" means.

Examples:
  • "Data fetched successfully and validated"
  • "Infrastructure deployed without errors"
  • "All quality checks passed"

Enter criteria or press Enter to skip:

6. Confirm Requirements

Display collected requirements and confirm:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Skill Requirements Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Name: {skill_name}
Description: {description}
Responsibility: {responsibility}
Tools: {tools}
Plugin: {plugin_name}
Location: plugins/{plugin_name}/skills/{skill_name}/

Inputs: {inputs}
Outputs: {outputs}

Optional:
  Handler Type: {handler_type or "Not a handler"}
  Workflow Steps: {steps or "Not specified"}
  Scripts: {scripts or "None specified"}
  Completion Criteria: {criteria or "Not specified"}

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Proceed with skill creation? (y/n):

Validation:

• Require explicit "y" or "yes" to proceed
• "n" or "no" cancels gracefully
• Any other input prompts again

Return Structure

After confirmation, return:

{
  "status": "success",
  "artifact_type": "skill",
  "requirements": {
    "name": "{skill_name}",
    "description": "{description}",
    "responsibility": "{responsibility}",
    "tools": "{tools}",
    "plugin_name": "{plugin_name}",
    "output_path": "plugins/{plugin_name}/skills/{skill_name}/",
    "handler_type": "{handler_type}",
    "inputs": "{inputs}",
    "outputs": "{outputs}",
    "workflow_steps": "{steps}",
    "scripts_needed": "{scripts}",
    "completion_criteria": "{criteria}"
  }
}

Error Handling

User Cancels: Return {"status": "cancelled", "message": "Skill creation cancelled by user"}

Invalid Input:

• Re-prompt up to 3 times
• After 3 failures, return error with validation message

Plugin Not Found: Return {"status": "error", "error": "Plugin not found: {plugin_name}", "step": "plugin_validation"}

Validation Rules

Skill Name

• Lowercase letters, numbers, hyphens only
• Must start with letter
• No consecutive hyphens
• 3-50 characters
• Examples: "data-fetcher", "terraform-deployer", "api-client"

Description

• 10-200 characters
• Starts with capital letter
• No newlines
• Clear and specific

Responsibility

• Non-empty
• Starts with lowercase
• Describes action (verb phrase)
• 20-300 characters

Tools

• From valid tool list: Bash, Read, Write, Edit, WebFetch, Grep, Glob, Skill
• Comma-separated
• At least one tool required

Plugin Name

• Must match existing plugin directory in plugins/
• Lowercase with hyphens
• Examples: "faber", "faber-cloud", "work"

Handler Type (if provided)

• Lowercase single word
• No spaces or special characters
• Examples: "iac", "hosting", "storage"

Inputs/Outputs

• Clear description of parameters/results
• Can be "none" for inputs if skill has no parameters

This workflow ensures all necessary information is collected before skill generation begins.