Skip to main content

​
GitHub App Setup

Whim connects to GitHub through a dedicated Whim GitHub App that provides secure, scoped access to your repositories — so agents can clone, branch, commit, and open PRs on your behalf.

​
Installing the GitHub App

The GitHub App is installed at the team level. Once installed, all workspaces in the team can access the granted repositories.
1

Open team settings

Navigate to your workspace → Settings → GitHub.
2

Install the GitHub App

Click Install GitHub App. You’ll be redirected to GitHub to authorize the Whim app.
3

Select repositories

Choose All repositories or select specific ones. You can change this anytime from GitHub App settings.
4

Confirm installation

After authorizing, you’re redirected back to Whim. Repositories become available in your workspaces.
You must be a team admin or owner to install or manage the GitHub App.

​
Repository Permissions

PermissionAccessPurpose
Repository contentsRead & writeClone repos, create branches, push commits
Pull requestsRead & writeCreate and update PRs from tasks
WebhooksReadReceive push, PR, and branch-delete events
MetadataReadList repositories and detect organization type
Modify repository access anytime from GitHub → Applications → Whim.

​
Linking Repos to Workspaces

  1. Open your workspace → Settings → Repositories.
  2. Select a repository from your team’s GitHub App installation.
  3. Tasks in the workspace automatically get access to the repo code.
Whim can also create new repositories directly from workspace settings — new repos are private by default.

​
Webhook Events

When code is pushed to a linked repository, Whim syncs the mirror so tasks always have the latest code. If the push modifies container-relevant files (e.g., Dockerfile, whim.json, package.json), Whim reconciles tracked branch images.
Whim tracks PRs on whim-* branches. Tracked actions: opened, reopened, closed, ready_for_review, converted_to_draft, synchronize. PR state is visible directly on the task in Whim.
When a branch is deleted on GitHub, Whim removes tracked branch state across workspaces using that repository.
All webhook payloads are verified using HMAC-SHA256 signatures for security.

​
Skills & Slash Commands

Skills are custom instructions that extend what your AI agents can do. Each skill is a Markdown file that gets injected into the agent’s context when invoked via a slash command (e.g., /code-review).

​
Anatomy of a Skill

---
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
---

Review the provided code changes for correctness, style,
performance, and security issues...

​
Frontmatter Fields

FieldDefaultDescription
nameauto-inferredSlash command identifier
descriptionauto-inferredShort description shown in the skill catalog
user_invocabletrueWhether users can invoke via slash command
argument_hint—Usage hint (e.g., <pr-number>)
allowed_tools—Comma-separated list of tools the skill can use
contextdefaultdefault runs inline; fork runs in isolated context
agent—Agent type when context is fork
model—Model override (e.g., claude-opus-4-6)
disable_model_invocationfalsePrevent agent from invoking this skill on its own

​
Built-in Skills

commit-message, fix-issue, pr-summary, api-design
code-review, refactor, test-generation, security-review
debug, document, explain-code
deploy-checklist, performance-audit

​
Creating Custom Skills

  1. Go to Settings → Skills.
  2. Click Add Skill and fill in the name, description, and instructions.
  3. Optionally set context, allowed_tools, model, and other frontmatter fields.
  4. Save — it’s immediately available as a slash command.
You can also import skills by uploading Markdown files or pasting Claude Code settings.json content. Multi-file skills (up to 50 files, 2 MB total) use a SKILL.md entry point.

​
Skill Scope & Resolution

Skills can be configured at workspace and user levels. Resolution order:
  1. Workspace skills (all enabled)
  2. Minus any the user has individually disabled
  3. Plus user’s personal skills (override workspace skills with the same name)
Enable built-in skills from the catalog in Settings → Skills — toggle on/off and optionally customize content.
ConstraintLimit
Skills per config100
Files per multi-file skill50
Total size per skill2 MB

​
Plugins

Plugins are published packages from plugin marketplaces that add entire tool ecosystems to your AI agents — integrations, language servers, code quality tools, and more.

​
How Plugins Work

Plugins use a name@marketplace format (e.g., github@claude-plugins-official). Enable a plugin and its tools become available to agents in your workspace.
Plugins are available for Claude and CCR providers. They are not currently supported for Codex.

​
Available Marketplaces

Claude Code Plugins

Anthropic’s first-party plugins — development workflows, testing tools, and agent capabilities.

Claude Plugins Official

Anthropic’s official registry — integrations, language servers, infrastructure tools, and more.
You can also define custom marketplaces pointing to GitHub repositories for private or org-specific plugins.

​
Plugin Catalog

frontend-design, code-review, commit-commands, e2e-testing, add-testids, code-simplifier, questions, codex, explore, plan
lsp-typescript, lsp-python, lsp-rust, lsp-go, lsp-java, lsp-csharp, lsp-ruby, lsp-php, lsp-swift, lsp-kotlin, lsp-html-css, lsp-yaml, lsp-json
github, slack, linear, jira, notion, confluence, google-drive, google-sheets, google-calendar, stripe, supabase, sentry, datadog, pagerduty
docker, kubernetes, terraform, aws, azure, gcp, vercel, netlify, cloudflare
eslint, prettier, stylelint, jest, vitest, pytest, snyk, sonarqube, dependabot
postgres, redis, mongodb

​
Enabling Plugins

  1. Go to Settings → Plugins in your workspace or user preferences.
  2. Search by name or filter by category.
  3. Toggle plugins on or off. Changes take effect for new tasks.

​
Plugin Scope & Resolution

Like skills, plugins merge across levels:
  1. Workspace plugins (all enabled)
  2. Minus any the user has disabled
  3. Plus user’s personal plugins

​
Custom Marketplaces

Add custom marketplace definitions in Settings → Plugins under advanced configuration: 
{
  "my-org-plugins": {
    "source": {
      "source": "github",
      "repo": "my-org/claude-plugins"
    }
  }
}
Reference plugins as plugin-name@my-org-plugins.
ConstraintLimit
Plugins per config100
Custom marketplaces50