Skip to main content

​
Overview

The /create-skill command helps you capture successful custom security testing approaches and save them as reusable specialist skills that can auto-load in future sessions.
This feature is in alpha - skills are created and saved, but auto-loading integration is not yet fully implemented.

​
When to Use

Save an approach as a skill after you’ve found success with:
  • Custom analysis focus (e.g., “focus on API security only”)
  • Custom priority ordering (e.g., “check auth before secrets”)
  • Specific testing methodologies that worked well
  • Domain-specific patterns (e.g., “mobile app security patterns”)

​
Usage

/create-skill
Claude will guide you through the skill creation process interactively.

​
Skill Creation Process

1

Capture Successful Approach

Describe what made your approach successful:
  • Custom priorities used
  • Specific focus areas
  • Testing techniques employed
  • Domain expertise applied
2

Define Skill Parameters

Specify how the skill should be used:
  • Name: Descriptive identifier (e.g., api_security_auth_focus)
  • Keywords: Trigger words for auto-loading
  • Domain: Target types where skill applies
3

Extract Reusable Patterns

Review and generalize the approach:
  • Identify generalizable patterns
  • Remove target-specific details
  • Document tool combinations
  • Define reusable priorities
4

Validate Token Budget

Ensure the skill fits within limits:
  • Maximum 500 tokens per skill
  • Warning at 10+ total custom skills
  • Consider consolidation if approaching limits

​
Skill Structure

Skills are saved to tiers/specialists/custom/[skill_name].md with this format: 
# [Skill Name]
# Created: [date]
# Source: Successful approach from [session]
# Token cost: [X] tokens
# Auto-loads: [trigger keywords]

## Core Philosophy
[What makes this approach unique/successful]

## When to Use
[Target types and situations]

## Approach
[Priorities, techniques, and tools]

## Integration with Python
[How this guides execution parameters]

​
Token Budget Guidelines

  • Minimum: 200 tokens (enough for useful content)
  • Recommended: 300-400 tokens (sweet spot)
  • Maximum: 500 tokens (hard limit, enforced)

​
Example

User: /create-skill

Claude: What successful approach should we save?

User: "We focused on API authentication, checking auth bypass before 
      anything else, and it found critical issues faster"

Claude:
  ✓ Extracting approach patterns...

  Skill preview:
  - Name: api_auth_priority
  - Keywords: API, REST, authentication, auth bypass
  - Focus: Authentication issues in web APIs
  - Priority: Auth bypass → API security → Input validation
  - Size: 380 tokens ✓

  Create this skill? [Y/n/Customize]

​
Quality Checks

Before saving a skill, ensure:
  • Not overfitted to one specific target
  • Token limit respected (under 500 tokens)
  • Keywords defined for auto-loading
  • Approach clearly documented
  • Integration with Python parameters explained

​
Managing Skills

Skills are stored in tiers/specialists/custom/ Common operations: 
# List all custom skills
ls tiers/specialists/custom/

# Disable a skill temporarily
mv skill_name.md skill_name.md.disabled

# Remove a skill
rm tiers/specialists/custom/skill_name.md

# Edit a skill
vim tiers/specialists/custom/skill_name.md
If you have 5+ custom skills, RAPTOR will prompt you quarterly to review usage stats and consider consolidation.

​
See Also

Creating Personas

Create custom expert personas

Extending RAPTOR

Add new framework capabilities

Build docs developers (and LLMs) love

Get started for freeTalk to us