Claude Code
Super­skills

What it solves

• Skills that don't trigger because the description is wrong
• Hookify rules that run on every message instead of the right one
• No examples of a complete, working skill to learn from
• No guide for when to use a skill vs a hook vs both

What's in the box

claude-code-architect/
├── SKILL.md          5 modes: CREATE CAPTURE IMPROVE HOOK AUDIT
├── README.md  INSTALL.md
├── references/
│   ├── skill-writing-guide.md  full anatomy
│   ├── hook-patterns.md        all events + regex ref
│   └── automation-strategies.md
└── examples/
    ├── example-pr-summary.md   complete reference skill
    └── example-hooks/          4 working hooks

What it solves

• Reviews that catch surface bugs but miss architectural issues
• Security checks that only cover the obvious OWASP top 10
• No confidence scores — just opinions with no priority signal
• Slow sequential review instead of parallel specialization

What's in the box

total-code-intelligence/
├── SKILL.md          QUICK FULL TARGETED PR-POST modes
├── references/
│   ├── confidence-scoring.md   0–100 rubric
│   └── security-patterns.md   injection → path traversal
└── agents/                    10 specialist agents
    ├── bug-hunter.md
    ├── security-auditor.md
    ├── type-design-analyzer.md
    ├── test-coverage-analyst.md
    └── + 6 more specialists

What it solves

• Four build surfaces, zero guidance on which to use when
• MCP docs explain what each thing is but not how to start one
• TypeScript and Python examples for the same patterns
• No working examples to copy and run immediately

What's in the box

claude-ecosystem-builder/
├── SKILL.md          paths A–E, discovery wizard
├── references/
│   ├── decision-matrix.md
│   ├── mcp-server-guide.md    TS + Python scaffolds
│   ├── agent-sdk-guide.md
│   └── full-stack-patterns.md
└── examples/
    ├── mcp-server-minimal/    Express + 2 tools (TS)
    ├── agent-sdk/typescript/  streaming agent
    ├── agent-sdk/python/
    └── mcpb-local/            bundled file reader

What it solves

• Every Next.js project starting with "clean and modern" and no real aesthetic direction
• Vercel env var setup that takes longer than building the feature
• shadcn setup, Neon Postgres, Clerk auth — documented but not assembled
• AI SDK patterns that work but require reading 5 docs pages

What's in the box

full-stack-launcher/
├── SKILL.md          SHIP DESIGN BUILD DEPLOY ENHANCE
├── references/
│   ├── design-system.md       14 laws + 8 aesthetic styles
│   ├── nextjs-patterns.md     App Router + Drizzle
│   ├── vercel-deploy.md       bootstrap → CI/CD
│   └── ai-features.md         AI SDK + tool calling
└── examples/
    ├── nextjs-starter/        landing page (Next.js 15)
    └── ai-chat-app/           streaming Claude chat