Generate UML diagrams from real code with a multi-stage audit pipeline that verifies business correctness, diagram-code consistency, and notation readability, backed by an adversarial diagram auditor. Also strengthens mutation test suites by writing outcome-asserting tests.
Use this agent in tier T1 of verify-then-draw to independently check that a diagram matches the real code CROSS-MODULE, finding nodes/transitions/actors/relations that are MISSING, EXTRA, or WRONG. Invoke after drawing or updating any code-derived diagram (state, use case, sequence, component, ERD). It reads code itself (does not trust the diagram author) and reports evidence with file:line. Does not edit anything.
Use this agent in tier T0 of verify-then-draw when a mutation run scores below the break threshold — to strengthen tests until the gate passes. It reads the project's mutation setup from PROJECT.yaml (runner-agnostic), writes OUTCOME-asserting tests that kill LIKELY-KILLABLE mutants, marks verified-equivalent mutants using the runner's disable syntax, and loops (re-run mutation → re-classify) until the score is ≥ break threshold or only equivalents remain. Does NOT change production logic — only tests + disable-comments with justification. Works for any mutation runner (Stryker, PIT, mutmut, Infection, cargo-mutants…) by reading tool-specifics from config rather than hardcoding them.
Uses power tools
Uses Bash, Write, or Edit tools
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
~/.claude/)Tổng hợp từ tài liệu chính thức Anthropic (code.claude.com, platform.claude.com), Engineering Blog, blog claude.com, MindStudio, ClaudeFast, và best practices cộng đồng GitHub. Áp dụng MỌI project. Tối ưu cho dev người Việt — tiếng Việt cho comment/log/commit, tiếng Anh chuẩn convention cho identifier.
👉 Lần đầu xem repo? Đọc INTRODUCTION.md cho overview ngắn 1 phút.
📅 Đã verify: 2026-05-16 vs Claude Code v2.1.142 + Opus 4.7. Model ID, version-gating, env var, slash command, hook event đã cross-check với docs Anthropic + raw CHANGELOG.
⚠️ Một số nội dung có thể outdated khi Anthropic ship version mới hoặc deprecate model. Phát hiện sai sót → submit issue/PR tại github.com/MinhThang1009/dotclaude/issues. Xem CONTRIBUTING.md cho quy trình. Lịch sử thay đổi: CHANGELOG.md.
~/.claude/ sau khi càiRepo source có thêm
docs/,.github/,scripts/,tests/,requirements-test.txt,.gitattributes,LICENSE,CHANGELOG.mdcho GitHub browsing + CI — KHÔNG copy vào~/.claude/(xem Bước 3).
~/.claude/
├── CLAUDE.md # Hướng dẫn global (load mọi session, ~79 dòng)
├── settings.json # Quyền, hooks, env vars
├── rules/
│ ├── communication.md # ✅ Auto-import (essential mỗi turn)
│ ├── security.md # ✅ Auto-import (essential mọi task)
│ ├── verification.md # ✅ Auto-import (verify & avoid past mistakes)
│ ├── coding-standards.md # ✅ Auto-import (coding conventions)
│ ├── git-workflow.md # ✅ Auto-import (git workflow)
│ └── plan-workflow.md # ✅ Auto-import (8 principles for any implementation plan)
├── plugins/ # 36 plugins (mỗi plugin có agents/ và/hoặc skills/ bên trong)
│ ├── pr-review-toolkit/ # PR review: code-reviewer, silent-failure-hunter, type-design-analyzer...
│ │ ├── .claude-plugin/
│ │ ├── agents/ # code-reviewer.md, code-simplifier.md, comment-analyzer.md...
│ │ └── skills/ # code-review/, full-review/, review-pr/
│ ├── feature-dev/ # Phát triển tính năng: code-architect, code-explorer
│ │ ├── .claude-plugin/
│ │ ├── agents/
│ │ └── skills/ # explain/, feature-dev/
│ ├── commit-commands/ # Git workflow: commit, commit-push-pr, clean-gone
│ │ ├── .claude-plugin/
│ │ └── skills/
│ ├── debug/ # Debug: debugger agent + /debug skill
│ ├── documentation/ # Docs: documentation-engineer, nextjs-developer
│ ├── test-toolkit/ # Test: test-writer, test-analyzer
│ ├── security-guidance/ # Security: security-auditor (OWASP)
│ ├── performance/ # Perf: performance-engineer, dependency-manager
│ ├── code-modernization/ # Architecture review: architecture-critic
│ ├── code-simplifier/ # Refactor: /refactor skill
│ ├── hookify/ # Tạo hooks: conversation-analyzer + /hookify skill
│ ├── frontend-design/ # UI: /frontend-design skill
│ ├── session/ # Session: /context-check, /handoff
│ ├── claude-md-management/ # Audit CLAUDE.md: /claude-md-management skill
│ └── plugin-dev/ # Phát triển plugin: 7 skills (agent/skill/hook/mcp/command/structure/settings)
npx claudepluginhub minhthang1009/dotclaude --plugin verify-then-drawAdds educational insights about implementation choices and codebase patterns (mimics the deprecated Explanatory output style)
Audit, evaluate quality, and improve CLAUDE.md files. Captures session learnings and keeps project memory up to date.
Build high-quality MCP servers using FastMCP or the MCP SDK
Build complex HTML artifacts (React, Tailwind, shadcn/ui) with state management for inline rendering in Claude
Interactive learning mode that requests meaningful code contributions at decision points (mimics the unshipped Learning output style)
Ultra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Unified capability management center for Skills, Agents, and Commands.