Skip to main content

​
Overview

Skills are custom instructions that extend what your AI agents can do. Each skill is a Markdown file with optional YAML frontmatter that defines a specialized workflow — like code review, commit message generation, or deployment checklists. Skills are invoked via slash commands (e.g., /code-review) directly in the task prompt or during a conversation with the agent.

​
How Skills Work

A skill is a Markdown document that gets injected into the agent’s context when invoked. The YAML frontmatter configures how the skill appears and behaves, while the Markdown body contains the actual instructions. 
---
name: code-review
description: Review code changes with detailed feedback
user_invocable: true
argument_hint: <pr-number>
allowed_tools: Read, Grep, Glob
context: fork
model: claude-sonnet-4-20250514
---

# Code Review

Review the provided code changes for:
- Correctness and potential bugs
- Code style and consistency
- Performance considerations
- Security issues
...

​
Frontmatter Fields

FieldTypeDefaultDescription
namestringauto-inferredDisplay name and slash command identifier
descriptionstringauto-inferredShort description shown in the skill catalog
user_invocablebooleantrueWhether users can invoke via slash command
argument_hintstring—Usage hint shown after slash command (e.g., <pr-number>)
allowed_toolsstring—Comma-separated list of tools the skill can use
contextstringdefaultdefault runs inline; fork runs in an isolated context
agentstring—Agent type to use when context is fork
modelstring—Specific model override (e.g., claude-opus-4-6)
disable_model_invocationbooleanfalsePrevent the agent from invoking this skill on its own

​
Multi-File Skills

For complex workflows, skills can contain multiple files. A multi-file skill must include a SKILL.md as the main entry point, plus any supporting files (helpers, templates, data). Multi-file skills support up to 50 files and 2 MB total.

​
Built-in Skills

Whim ships with a catalog of built-in skills you can enable with one click. These are organized into categories:
SkillDescription
commit-messageGenerate conventional commit messages from staged changes
fix-issueFix a GitHub issue by number
pr-summarySummarize pull request changes for review
api-designDesign RESTful API endpoints with types and validation
SkillDescription
code-reviewReview code changes with quality and bug feedback
refactorRefactor code for clarity and maintainability
test-generationGenerate unit and integration tests
security-reviewReview code for security vulnerabilities
SkillDescription
debugSystematic debugging workflow
documentGenerate documentation for code
explain-codeExplain code with clear breakdowns
SkillDescription
deploy-checklistPre-deployment verification checklist
performance-auditAnalyze code for performance issues

​
Creating Custom Skills

You can create your own skills from the Settings UI:
1

Open skill settings

Go to Settings > Skills in your workspace or user preferences.
2

Create a new skill

Click Add Skill and fill in the name, description, and instructions using the structured editor.
3

Configure advanced options

Optionally set context, allowed_tools, model, and other frontmatter fields to fine-tune how the skill runs.
4

Save

Save the skill. It immediately becomes available as a slash command in your tasks.
You can also import skills by uploading Markdown files, dragging and dropping skill directories, or pasting Claude Code settings.json content.

​
Using Slash Commands

Invoke any enabled skill as a slash command in the task composer or during a conversation: 
/code-review
/fix-issue 42
/deploy-checklist
Skills with argument_hint configured will show usage hints when you start typing the command. The argument is passed as context to the skill instructions.

​
Skill Scope

Skills can be configured at two levels:

​
Workspace Skills

  • Defined by workspace admins in Settings > Workspace Defaults > Skills
  • Available to all workspace members by default
  • Members can individually disable workspace skills they don’t want

​
User Skills

  • Defined in Settings > User Defaults > Skills
  • Available across all of your workspaces
  • Override workspace skills if they share the same name

​
Resolution Order

When a task runs, skills are resolved by merging:
  1. Workspace skills — all enabled skills from workspace defaults
  2. Minus disabled — remove any workspace skills the user has disabled
  3. Plus user skills — merge in the user’s personal skills (user skills override workspace skills with the same name)

​
Enabling Catalog Skills

To enable built-in skills from the catalog:
  1. Go to Settings > Skills.
  2. Browse or search the catalog by name, description, or category.
  3. Toggle skills on or off. Enabled skills are immediately available as slash commands.
  4. Optionally customize a catalog skill’s content after enabling — your modifications are saved as an override.

​
Limits

ConstraintLimit
Skills per workspace/user config100
Files per multi-file skill50
Total size per skill2 MB
Skill name characters1–100, alphanumeric, hyphens, and underscores