Skills, sub-agents, and playbooks for Claude Code, Cursor, and any Agent Skills–compatible AI coding assistant.
What you get: 30+ curated skills, 19 sub-agents, 7 team role personas, and operational playbooks - all installable individually or as bundles, based on the open Agent Skills standard. Turn your AI coding assistant from a generalist into a domain specialist.
See CHANGELOG.md for release notes.
Table of Contents
# 1. Install (interactive - pick what you want)
npx skills add krzysztofsurdy/code-virtuoso
# 2. Try a skill immediately
/ticket-writer
# Follow the prompts - paste the output into Jira, Linear, or GitHub Issues.That is the whole tour. For fine-grained install modes (per-skill, global, offline, auto-update) see Installation (detailed).
- Battle-tested, not LLM-generated. Every skill was researched against official docs and community best practices, then hand-curated. No hallucinated API references.
- Progressive disclosure keeps your context clean. Skills load a short
SKILL.mdfirst and only pull in deeper references when the task calls for them. Less context rot, sharper answers. - One installation, many tools. Built on the open Agent Skills standard - works with Claude Code, Cursor, Windsurf, Copilot, Codex, Gemini CLI, and more. No lock-in.
- Team-shaped, not pattern-shaped. The library ships 7 role personas (Product Manager, Architect, Backend Dev, Frontend Dev, QA, PM, Scrum Master) and 15 sub-agents that delegate like a real team rather than a single over-prompted assistant.
| Category | Count | Summary |
|---|---|---|
| Knowledge | 18 | Principles, craft, process, and agent workflow references |
| Tools | 11 | Interactive generators - rules files, tickets, agents, plugins, PR messages, reports, updates, screenshots, image annotation |
| Frameworks | 4 | Symfony, Django, LangChain component libraries and upgrade guides |
| Playbooks | 5 | Step-by-step procedures for recurring operational tasks |
| Roles | 6 | Team-role reference skills used by role agents |
| Agents | 19 | 12 specialist + 7 role sub-agents with tool, isolation, and memory metadata |
| Teams | 3 | Pre-composed agent teams with bundled skills and coordination protocols |
Grouped by theme. Each skill is a self-contained reference with a short SKILL.md and deep-dive files under references/.
| Skill | Summary |
|---|---|
| Design Patterns | 26 Gang of Four patterns with idiomatic implementations |
| SOLID | All five SOLID principles with multi-language examples |
| Clean Architecture | Clean/Hexagonal Architecture and DDD fundamentals |
| Refactoring | 67 refactoring techniques and 22 code smells |
| Skill | Summary |
|---|---|
| Testing | Testing pyramid, TDD schools, test doubles, strategies |
| Debugging | Systematic debugging methodology and post-mortem templates |
| API Design | REST and GraphQL design principles and evolution strategies |
| Database Design | Schema modeling, indexing strategies, migration patterns |
| Performance | Profiling, caching, database optimization, N+1 prevention |
| Security | OWASP Top 10, auth patterns, secure coding practices |
| Accessibility | WCAG compliance, ARIA patterns, keyboard navigation, a11y testing |
| Skill | Summary |
|---|---|
| Scrum | Sprint goals, events, roles, and facilitation templates |
| Git Workflow | Branching strategies, commit conventions, PR patterns |
| CI/CD | Pipeline design, deployment strategies, environment promotion |
| Microservices | Saga, CQRS, event sourcing, circuit breakers, service mesh |
| Skill | Summary |
|---|---|
| Verification Before Completion | Evidence-based completion discipline and tiered definition of done |
| Dispatching Parallel Agents | Fan-out/fan-in patterns, subagent briefing, result synthesis |
| Subagent-Driven Development | One-fresh-agent-per-task execution with two-stage review gates |
Interactive generators. Each is user-invocable via /skill-name.
| Skill | Summary |
|---|---|
| Agentic Rules Writer | Generate rules files for Claude Code, Cursor, Windsurf, Copilot, Gemini, Roo Code, or Amp |
| Ticket Writer | Write tickets of the right type - story, subtask, issue, bug, epic, or initiative |
| Agent Creator | Design a sub-agent with proper frontmatter, tool permissions, isolation, and system prompt |
| Plugin Creator | Scaffold a complete Claude Code plugin - manifest, skills, agents, hooks, MCP/LSP servers |
| Brainstorming | Pre-implementation design exploration - turn a vague idea into an approved spec |
| Using Ecosystem | Guided tour and discovery advisor for the ecosystem |
| PR Message Writer | Write structured pull request messages with technical documentation and testing instructions |
| Report Writer | Generate standalone HTML reports summarizing changes, investigations, or architectural decisions |
| Stakeholder Update Writer | Draft stakeholder Slack or email updates about project status, blockers, and decisions |
| Web Screenshot | Capture web page screenshots with shot-scraper - full page, element, viewport, multi-device, batch |
| Image Annotate | Mark up images with arrows, boxes, text, numbered steps, highlights, blur, and solid redaction |
| Skill | Summary |
|---|---|
| Symfony Components | 38 Symfony components for PHP 8.3+ and Symfony 7.x |
| Symfony Upgrade | Deprecation-first upgrade guide for minor and major versions |
| Django Components | 33 Django components for Python 3.10+ and Django 6.0 |
| LangChain Components | 17 LangChain references - models, agents, tools, retrieval, LangGraph |
| Skill | Summary |
|---|---|
| PHP Upgrade | PHP version upgrade process with Rector, PHPCompatibility, breaking changes |
| Composer Dependencies | Safe dependency updates, security auditing, automated update tools |
| Finishing Branch | End-to-end branch finishing - pre-push checks, PR messages, cleanup, recovery |
| Ticket Delivery | End-to-end ticket delivery - analysis, investigation, planning, TDD implementation, commit, and PR |
| Worktree Ops | Create, list, switch, and remove git worktrees for parallel development sessions |
| Skill | Summary |
|---|---|
| Product Manager | Requirements gathering, PRD writing, prioritization, acceptance criteria |
| Architect | System design, component boundaries, API contracts, ADRs |
| Backend Dev | API implementation, data models, TDD workflows |
| Frontend Dev | UI components, accessibility, state management |
| QA Engineer | Test planning, test design, bug reporting, release sign-off |
| Project Manager | PRINCE2-based delivery, risk management, progress tracking |
Sub-agents follow the Claude Code sub-agents standard. Specialist agents handle one focused task type. Role agents embody a team position and carry persistent project memory. See AGENTS.md for full specifications, delegation patterns, and chaining examples.
| Kind | Agent | Tools | Isolation | Memory | Purpose |
|---|---|---|---|---|---|
| Specialist | Investigator | Read, Grep, Glob, Bash | - | - | Deep codebase exploration, dependency mapping |
| Specialist | Implementer | All | worktree | - | TDD red-green-refactor execution |
| Specialist | Reviewer | Read, Grep, Glob, Bash | - | - | Structured code review (SOLID, OWASP, smells) |
| Specialist | Refactor Scout | Read, Grep, Glob, Bash | - | - | Code smell scanning, complexity hotspots |
| Specialist | Dependency Auditor | Bash, Read, Grep, Glob | - | - | CVE checks, outdated packages, license audit |
| Specialist | Doc Writer | Read, Grep, Glob, Bash, Write, Edit | - | - | Changelogs, API docs, migration guides |
| Specialist | Migration Planner | Read, Grep, Glob, Bash | - | - | Migration safety analysis, rollback paths |
| Specialist | Test Gap Analyzer | Read, Grep, Glob, Bash | - | - | Missing test coverage, untested edge cases |
| Specialist | Cold Reviewer | Read, Grep, Glob, Bash | - | - | Zero-context code review, fresh-eyes findings |
| Specialist | Acceptance Verifier | Read, Grep, Glob, Bash | - | - | Spec compliance checking, criteria coverage matrix |
| Specialist | Readiness Checker | Read, Grep, Glob, Bash | - | - | Pre-implementation readiness gate |
| Specialist | Course Corrector | Read, Grep, Glob, Bash | - | - | Mid-workflow change impact analysis |
| Role | Product Manager | Read, Grep, Glob, Bash | - | project | Requirements, PRDs, prioritization |
| Role | Architect | Read, Grep, Glob, Bash | - | project | System design, ADRs, trade-offs |
| Role | Backend Dev | Read, Edit, Write, Bash, Grep, Glob | worktree | - | API implementation, data models, TDD |
| Role | Frontend Dev | Read, Edit, Write, Bash, Grep, Glob | worktree | - | UI components, accessibility, state |
| Role | QA Engineer | Read, Grep, Glob, Bash | - | project | Test plans, bug reports, release sign-off |
| Role | Project Manager | Read, Grep, Glob, Bash | - | project | PRINCE2 stages, risk, progress tracking |
| Role | Scrum Master | Read, Grep, Glob, Bash | - | - | Sprint planning, goals, retrospectives |
Pre-composed agent teams with bundled skills and coordination protocols. Pick a team instead of assembling agents manually. See spec/team-spec.md for the format.
| Team | Lead | Agents | Workflow | Use case |
|---|---|---|---|---|
| Development Team | Product Manager | PM, Architect, Backend Dev, Frontend Dev, QA | Hybrid | Full feature delivery from requirements to merged PR |
| Review Squad | Reviewer | Reviewer, Cold Reviewer, Acceptance Verifier | Parallel | Multi-perspective code review with triaged findings |
| War Room | Architect | Architect, PM, Backend Dev, QA | War Room | Structured debate for high-stakes technical decisions |
# Pick individual skills
npx skills add krzysztofsurdy/code-virtuoso --skill design-patterns --skill refactoring
# Everything at once
npx skills add krzysztofsurdy/code-virtuoso --all
# Install globally (available in every project)
npx skills add krzysztofsurdy/code-virtuoso -g
# Preview what's available without installing
npx skills add krzysztofsurdy/code-virtuoso --listnpx skills check # see what can be updated
npx skills update # update everything installedmacOS / Linux (zsh)
echo '_skills_marker="${TMPDIR:-/tmp}/.skills-updated-$(date +%Y%m%d)"
[ ! -f "$_skills_marker" ] && (npx skills update --yes >/dev/null 2>&1 && touch "$_skills_marker" &)' >> ~/.zshrcWindows (PowerShell)
Add-Content $PROFILE '$marker = "$env:TEMP\.skills-updated-$(Get-Date -Format yyyyMMdd)"; if (-not (Test-Path $marker)) { Start-Job { npx skills update --yes *> $null; New-Item $using:marker -Force } | Out-Null }'Project-level (git post-merge hook)
printf '#!/bin/sh\nnpx skills update --yes >/dev/null 2>&1 &\n' > .git/hooks/post-merge
chmod +x .git/hooks/post-mergeCompanion tools that pair naturally with this library.
Distributed, git-backed graph issue tracker that gives AI agents persistent, structured memory for long-horizon tasks. Replaces ad-hoc markdown planning files with a dependency-aware task graph.
github.com/gsd-build/get-shit-done
Meta-prompting and context engineering system for Claude Code, OpenCode, Gemini CLI, and Codex. Targets context rot with spec-driven development, subagent orchestration, and state management.
github.com/agentika-labs/grepika
MCP server that replaces built-in grep/file search with ranked, compact results using roughly 80% fewer tokens. Combines FTS5 full-text search, parallel grep, and trigram indexing with BM25 ranking.
Which AI coding assistants does this support? Anything that follows the Agent Skills open standard, including Claude Code, Cursor, Windsurf, GitHub Copilot, Codex, Gemini CLI, Cline, OpenCode, Continue, Trae, Roo Code, Amp, and more. Skills use portable frontmatter; platform-specific extensions (tool names, isolation, memory) are noted per-agent.
Do I have to install the full bundle?
No. Install only the skills you need with --skill <name>, or install a single plugin (e.g., role-backend-dev, knowledge-virtuoso, tool-ticket-writer). Each plugin is documented in .claude-plugin/marketplace.json.
How do I update?
Run npx skills update. For hands-free upkeep, use one of the auto-update snippets above.
Can I use skills across multiple projects?
Yes - install with -g for global availability. Local installs live in the current project only.
Can I contribute my own skills or agents?
Yes. See CONTRIBUTING.md for conventions, the research workflow, marketplace.json update rules, and version-bumping policy.
Why not just prompt-engineer one long CLAUDE.md file? One long rule file consumes context on every request. Skills use progressive disclosure - metadata sits in context, bodies load only when triggered, deep references load only when explicitly needed. You get specialist knowledge without paying for it until it's relevant.
See CONTRIBUTING.md for:
- How to add a new skill or agent
- Naming conventions and frontmatter rules
- Marketplace configuration and plugin tiers
- Version-bumping policy
- Quality standards and language/provider-agnosticism rules
Commit conventions are in CLAUDE.md. No AI co-author lines.
MIT - see LICENSE.
Click to expand
code-virtuoso/
├── agents/ # Sub-agent definitions (Anthropic standard)
│ ├── investigator.md # Specialist: codebase exploration
│ ├── implementer.md # Specialist: TDD execution
│ ├── reviewer.md # Specialist: code review
│ ├── refactor-scout.md # Specialist: code smell scanning
│ ├── dependency-auditor.md # Specialist: CVE/license audit
│ ├── doc-writer.md # Specialist: documentation generation
│ ├── migration-planner.md # Specialist: migration safety
│ ├── test-gap-analyzer.md # Specialist: coverage gaps
│ ├── cold-reviewer.md # Specialist: zero-context review
│ ├── acceptance-verifier.md # Specialist: criteria compliance
│ ├── readiness-checker.md # Specialist: pre-implementation gate
│ ├── course-corrector.md # Specialist: change impact analysis
│ ├── product-manager.md # Role: requirements, PRDs
│ ├── architect.md # Role: system design, ADRs
│ ├── backend-dev.md # Role: API implementation
│ ├── frontend-dev.md # Role: UI components
│ ├── qa-engineer.md # Role: test planning
│ ├── project-manager.md # Role: delivery management
│ └── scrum-master.md # Role: sprint facilitation
├── skills/
│ ├── knowledge/
│ │ ├── api-design/
│ │ ├── clean-architecture/
│ │ ├── debugging/
│ │ ├── design-patterns/
│ │ ├── refactoring/
│ │ ├── microservices/
│ │ ├── cicd/
│ │ ├── git-workflow/
│ │ ├── accessibility/
│ │ ├── database-design/
│ │ ├── performance/
│ │ ├── scrum/
│ │ ├── security/
│ │ ├── solid/
│ │ ├── testing/
│ │ ├── verification-before-completion/
│ │ ├── dispatching-parallel-agents/
│ │ └── subagent-driven-development/
│ ├── roles/
│ │ ├── product-manager/
│ │ ├── architect/
│ │ ├── backend-dev/
│ │ ├── frontend-dev/
│ │ ├── qa-engineer/
│ │ └── project-manager/
│ ├── frameworks/
│ │ ├── django/
│ │ │ └── django-components/
│ │ ├── langchain/
│ │ │ └── langchain-components/
│ │ └── symfony/
│ │ ├── symfony-components/
│ │ └── symfony-upgrade/
│ ├── playbooks/
│ │ ├── php-upgrade/
│ │ ├── composer-dependencies/
│ │ ├── finishing-branch/
│ │ ├── ticket-delivery/
│ │ └── worktree-ops/
│ └── tools/
│ ├── agentic-rules-writer/
│ ├── ticket-writer/
│ ├── agent-creator/
│ │ └── phases/ # Guided phase files
│ ├── agent-teams/
│ │ └── teams/ # Bundled team library (development-team, review-squad, war-room)
│ ├── dispatching-agent-teams/ # Team dispatcher (peer/sequential/inline modes)
│ ├── plugin-creator/
│ │ └── phases/ # Guided phase files
│ ├── brainstorming/
│ │ └── phases/ # Guided phase files
│ ├── using-ecosystem/
│ ├── pr-message-writer/
│ ├── report-writer/
│ ├── stakeholder-update-writer/
│ ├── web-screenshot/
│ └── image-annotate/
├── spec/ # Format specifications
│ ├── skill-spec.md
│ ├── agent-spec.md
│ ├── plugin-spec.md
│ └── team-spec.md
├── template/ # Starter templates
│ ├── skill-template.md
│ ├── agent-template.md
│ └── team-template.md
├── AGENTS.md
├── CHANGELOG.md
├── CONTRIBUTING.md
├── LICENSE
└── README.md