The catalog was Java-shaped and per-line only: a diff could satisfy every rule and still force edits in eleven files. Three new reference layers cover what rules.md is blind to by construction. References: - fowler.md: twelve change-shape smells, each with a detection question, a TypeScript remedy, and a "when NOT to fix" case - typescript.md: TS1-TS10 native smells (any leakage, assertion vs validation, exhaustiveness, strictness flags), plus the G23-vs-discriminated-union resolution - tooling.md: which smells a tool decides. knip replaces the archived ts-prune; documents that no-unsafe-type-assertion ships in no preset, not even strict-type-checked Review posture, sourced to Google eng-practices and Conventional Comments: approve when the change improves code health; a correct finding nobody acts on is a false positive; polish is labelled ignorable. Every catalog entry now carries severity, when-NOT-to-fix, and mechanically-detected-by. Contested rules marked as positions with counter-arguments rather than law: function length (Ousterhout's deep modules), comments, G23 polymorphism. Refactoring workflow gains step 3, a Fowler sweep over all twelve that picks the root smell before planning; later steps shift to 4-8. SKILL.md rewritten as a router, 189 -> 108 lines: sharpened description, one source of truth per concept, completion criteria per step, and ~154 lines of frontmatter restatement removed across the step files. Includes the research run the changes are sourced from under _bmad-output/. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
BMAD Skills Marketplace
A collection of skills for Claude Code that extend its capabilities with specialized knowledge and workflows.
What are Skills?
Skills are self-contained instruction sets that teach Claude Code how to perform specialized tasks. Each skill is defined in a SKILL.md file with optional supporting resources like scripts, templates, and reference documentation.
Repository Structure
<root>/
├── skills/ # All skill implementations
│ └── [skill-name]/ # Individual skill folder
│ ├── SKILL.md # Required - Main skill definition
│ ├── scripts/ # Optional - Supporting scripts
│ ├── references/ # Optional - Reference documentation
│ └── assets/ # Optional - Images, resources
├── template/ # Starter template for new skills
├── spec/ # Skills specification documentation
├── CONTRIBUTING.md # Contribution guidelines
└── README.md # This file
Installation
Method 1: Install via skills CLI (Recommended)
The easiest way to install skills using the skills package. Supports Claude Code, Cursor, Codex, OpenCode, and 20+ other agents.
# Install all skills
npx skills add bmad-labs/skills
# Install specific skills for TypeScript development
npx skills add bmad-labs/skills --skill typescript-clean-code --skill typescript-e2e-testing --skill typescript-unit-testing
# Install to specific agent (e.g., claude-code, cursor, codex)
npx skills add bmad-labs/skills -a claude-code
# Install globally (user-level, available in all projects)
npx skills add bmad-labs/skills -g
Method 2: Install via Plugin Marketplace
This method installs the entire plugin with all skills automatically detected.
Step 1: Add the marketplace
/plugin marketplace add bmad-labs/skills
Step 2: Install the plugin
/plugin install bmad-labs@skills
Step 3: Verify installation
# Open the plugin manager to see installed plugins
/plugin
# Or ask Claude: "What skills are available?"
Method 3: Install via settings.json
Add to your project's .claude/settings.json:
{
"extraKnownMarketplaces": {
"bmad-labs": {
"source": {
"source": "github",
"repo": "bmad-labs/skills"
}
}
},
"enabledPlugins": {
"bmad-labs@skills": true
}
}
Method 4: Install Individual Skills
If you only need specific skills, add them individually to your project's .claude/settings.json:
{
"skills": [
"github:bmad-labs/skills/skills/typescript-clean-code",
"github:bmad-labs/skills/skills/typescript-e2e-testing",
"github:bmad-labs/skills/skills/typescript-unit-testing"
]
}
Available skill paths:
github:bmad-labs/skills/skills/mcp-buildergithub:bmad-labs/skills/skills/typescript-e2e-testinggithub:bmad-labs/skills/skills/typescript-unit-testinggithub:bmad-labs/skills/skills/skill-creatorgithub:bmad-labs/skills/skills/typescript-clean-codegithub:bmad-labs/skills/skills/book-convertergithub:bmad-labs/skills/skills/skill-from-bookgithub:bmad-labs/skills/skills/slides-generatorgithub:bmad-labs/skills/skills/ai-multimodalgithub:bmad-labs/skills/skills/atlassian-restgithub:bmad-labs/skills/skills/multi-repo-git-opsgithub:bmad-labs/skills/skills/trade-off-analysisgithub:bmad-labs/skills/skills/manual-testinggithub:bmad-labs/skills/skills/load-docsgithub:bmad-labs/skills/skills/rca-reportgithub:bmad-labs/skills/skills/jira-to-local
Troubleshooting
If skills are not detected after installation:
-
Clear the plugin cache:
rm -rf ~/.claude/plugins/cache -
Restart Claude Code (exit and run
claudeagain) -
Verify skills are loaded:
- Ask Claude: "What skills are available?"
- Or run
/pluginand check the Errors tab
-
Check plugin structure:
claude plugin validate .
Using Skills
Invoking a Skill
Once installed, you can invoke skills using the slash command format:
/skill-name [arguments]
Or simply describe what you need - Claude will automatically use relevant skills based on their descriptions.
Creating New Skills
- Copy the
template/folder toskills/your-skill-name/ - Edit
SKILL.mdwith your skill definition - Add any supporting scripts or references
- Test the skill locally
- Submit a PR
See CONTRIBUTING.md for detailed guidelines.
Skill Categories
| Category | Description |
|---|---|
| Development | Coding, architecture, testing, DevOps |
| Productivity | Documents, project management, workflows |
| Creative | Design, content creation, branding |
| Data | Processing, analysis, visualization |
Available Skills
| Skill | Description | Category |
|---|---|---|
| mcp-builder | Guide for creating high-quality MCP servers for LLM integrations | Development |
| typescript-e2e-testing | Comprehensive E2E testing for TypeScript/NestJS with Kafka, PostgreSQL, MongoDB, Redis | Development |
| typescript-unit-testing | Unit testing for TypeScript/NestJS with Jest, DeepMocked, mongodb-memory-server, pg-mem, Kafka, Redis | Development |
| typescript-clean-code | Clean Code principles and workflows for TypeScript development | Development |
| book-converter | Convert EPUB books to formatted Markdown | Productivity |
| skill-from-book | Convert book content into structured Claude Code skills | Development |
| slides-generator | Generate interactive presentation slides using React + Tailwind, export to PDF or HTML | Productivity |
| slide-maker | Design-system-agnostic slide/deck maker — brainstorm, generate, and export to editable PPTX, image PPTX, standalone HTML, or PDF; 34 token-driven layouts, bundled neutral theme, works with your design system or a suggested one | Productivity |
| ai-multimodal | Process and generate multimedia content using Google Gemini API (audio, images, videos, documents) | Data |
| atlassian-rest | Jira & Confluence integration via REST APIs — tickets, epics, pages, status reports, triage, bidirectional BMAD document sync | Productivity |
| multi-repo-git-ops | Git operations across multi-repo systems with submodules — branching, committing, pushing, syncing across parent and service repos | Development |
| release-please | Automated versioning and releases with Google's release-please — pipeline setup, Conventional Commits, pre-releases, monorepos | Development |
| trade-off-analysis | Structured technology trade-off analysis with scored comparison matrices for architecture decisions, build-vs-buy, and technology selection | Productivity |
| manual-testing | Plan, write, review, execute, and maintain manual test cases — API/backend, frontend, pipeline, AI/LLM, and infrastructure testing with risk-based prioritization | Development |
| load-docs | Load documents (files, folders, URLs, PDFs) fully into the main agent's context for follow-up Q&A, summarization, or review | Productivity |
| rca-report | Investigate and document production incidents — guides evidence collection during the investigation and produces rich, reproducible Root Cause Analysis reports with layered root cause, workaround playbook, Five Whys, and binned recommendations | Development |
| software-research | STORM-style, dev-specialized research: auto-detects a mode (library-eval, deep-research, architecture/ADR, spike, migration), runs expert lenses in parallel, and verifies every claim against primary sources (docs, RFCs, OSV/CVE, benchmarks) version-aware | Development |
| jira-to-local | Pull a Jira issue onto disk as readable local files — fields, comments, subtasks, worklogs, images, dev panel, and any linked Confluence page; read-only, markdown or JSON output | Productivity |
Recommended Skills from Other Repositories
Expand your Claude Code capabilities with these excellent skills from the community:
Vercel Labs Skills
# Install agent-browser skill for browser automation
npx skills add vercel-labs/agent-browser
# Install React best practices skill
npx skills add vercel-labs/agent-skills --skill vercel-react-best-practices
# Install all recommended skills at once
npx skills add vercel-labs/agent-browser && \
npx skills add vercel-labs/agent-skills --skill vercel-react-best-practices
| Repository | Skill | Description |
|---|---|---|
| vercel-labs/agent-browser | agent-browser | Browser automation and web interaction capabilities |
| vercel-labs/agent-skills | vercel-react-best-practices | React development best practices and patterns |
License
MIT License - See individual skills for specific licensing.