Skrrt Plugins

Marketplace overview for the skrrt-sh plugin catalog, installation flow, and bundled skills.

Table of Contents

• Installation
• Plugins
• Requirements
• Repository Structure
• License

Claude Code plugin marketplace by skrrt-sh — documentation, developer workflows, and productivity tools.

Installation

bunx skills add skrrt-sh/skills

Or from inside Claude Code:

/install skrrt-sh/skills

Team Setup

Add to your project's .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "skrrt-plugins": {
      "source": {
        "source": "github",
        "repo": "skrrt-sh/skills"
      }
    }
  }
}

Plugins

md-writer

Write well-structured markdown with YAML frontmatter, Mermaid diagrams, and markdownlint validation.

/plugin install md-writer@skrrt-plugins

Features:

• YAML frontmatter with required metadata fields
• Mermaid-only diagrams — no ASCII art
• Markdownlint compliance — 20+ rules enforced, zero violations
• PostToolUse hook — auto-validates .md files on Write/Edit
• Custom config — respects project-level .markdownlint.json

Usage:

/md-writer API integration guide for the payments service

Or just ask Claude to write markdown — the skill activates automatically.

See plugins/md-writer/ for full details.

Custom lint config: The plugin ships with a bundled default config. To override it, place your own config at your project root:

# Any of these will take precedence over the plugin default:
.markdownlint.json
.markdownlint.jsonc
.markdownlint.yaml
.markdownlint.yml

The validation hook walks up from the markdown file looking for the nearest config. If none exists, it falls back to the plugin's bundled default. The skill is written to conform to the same defaults.

ship

Create conventional commits with gitmojis, open focused PRs or MRs, and prepare release text with the matching forge CLI.

/plugin install ship@skrrt-plugins

Features:

• Uses the vivaxy/vscode-conventional-commits commit shape
• Uses upstream conventional commit type titles and descriptions
• Uses the same gitmoji dataset version referenced by that repo
• Splits work into dedicated commit, pr, release, and setup skills
• Detects whether the repo is hosted on GitHub or GitLab before choosing gh or glab
• Bundles skill-local forge-detection scripts for portable execution
• Documents a conservative git command subset for status, staging, commit, push, and release workflows
• Publishes review requests and releases with explicit non-interactive CLI commands
• Updates an existing CHANGELOG.md during release work when the repository has one
• Keeps the skills user-invocable and model-invocable
• Bundles a recommended Claude Code permissions template with ask rules for writes and force-push variants

Recommended first step — run /setup:

For the best experience, run /setup in your project before using the other skills. The setup skill wires directives into your project's CLAUDE.md (or AGENTS.md) so that Claude automatically uses the ship skills whenever it commits, opens a PR/MR, or prepares a release — no slash command needed.

Setup also configures a branching strategy (GitHub Flow, Trunk-Based Development, or Gitflow) so that all skills respect the correct target branches, merge rules, and release workflows.

/setup wire skrrt skills into this project

Without /setup, the commit, PR, and release skills will prompt you to run /setup first so that a branching strategy is configured.

Usage:

/commit prepare a clean conventional commit for the auth refresh-token changes
/pr open a review request for the auth refresh-token branch
/release draft release notes for v1.4.0

Or — after running /setup — just ask Claude to commit work, open a PR or MR, or prepare a release and it will use the skills automatically.

Recommended permissions:

Claude Code permissions are configured in .claude/settings.json, not in SKILL.md. This plugin includes a recommended template at plugins/ship/templates/claude-settings.json. Merge it into your project's .claude/settings.json if you want:

• read-only git and view commands allowed automatically
• mutating git, gh, and glab commands escalated with permissions.ask
• force-push variants escalated to human approval instead of silently allowed
• destructive commands such as git reset --hard denied

Evaluations:

This plugin also includes lightweight evaluation fixtures under plugins/ship/evals to support the Anthropic recommendation to test skills against representative scenarios before sharing them.

Requirements

• Claude Code v1.0.33+
• Node.js 18+
• jq (for validation hooks)

Repository Structure

.
├── .claude-plugin/
│   └── marketplace.json           # Marketplace manifest
├── plugins/
│   ├── md-writer/                 # Markdown writer plugin
│   │   ├── .claude-plugin/
│   │   │   └── plugin.json
│   │   ├── config/
│   │   │   └── markdownlint-default.json
│   │   ├── evals/
│   │   │   └── evals.json
│   │   ├── hooks/
│   │   │   ├── hooks.json
│   │   │   └── validate-md.sh
│   │   ├── skills/
│   │   │   └── md-writer/
│   │   │       └── SKILL.md
│   │   ├── package.json
│   │   └── package-lock.json
│   └── ship/                      # Commit, PR or MR, and release workflow plugin
│       ├── .claude-plugin/
│       │   └── plugin.json
│       ├── evals/
│       │   ├── commit-basic.json
│       │   ├── evals.json
│       │   ├── pr-github.json
│       │   └── release-changelog.json
│       ├── templates/
│       │   └── claude-settings.json
│       └── skills/
│           ├── commit/
│           │   ├── SKILL.md
│           │   ├── evals/
│           │   │   └── trigger-evals.json
│           │   └── reference/
│           │       ├── commit-types.md
│           │       └── gitmojis.md
│           ├── pr/
│           │   ├── SKILL.md
│           │   ├── evals/
│           │   │   └── trigger-evals.json
│           │   └── scripts/
│           │       └── detect-forge-cli.sh
│           ├── release/
│           │   ├── SKILL.md
│           │   ├── evals/
│           │   │   └── trigger-evals.json
│           │   └── scripts/
│           │       └── detect-forge-cli.sh
│           └── setup/
│               ├── SKILL.md
│               ├── evals/
│               │   └── trigger-evals.json
│               └── reference/
│                   └── branching-strategies.md
├── README.md
├── LICENSE
├── skills-lock.json
└── .gitignore

Contributing

Dev Setup

Install the pinned development skills before working on the plugins:

# Install dev skills (one-time, or after pulling new lock entries)
bunx skills add anthropics/skills --skill skill-creator

This restores .agents/ with the skill-creator used for eval workflows.

Running Evals

Use the skill-creator to test changes against the eval suites:

/skill-creator audit our skills, run evals

Eval workspaces (md-writer-workspace/, ship-workspace/) are gitignored — they are runtime artifacts from running evals, not committed. The eval definitions themselves live in plugins/*/evals/.

Project Layout for Dev Files

.agents/                  # Installed dev skills (gitignored, restored from lockfile)
skills-lock.json          # Lockfile for dev skills (committed)
md-writer-workspace/      # md-writer eval artifacts (gitignored)
ship-workspace/           # ship eval artifacts (gitignored)

License

MIT