tech-snacks

Tech Snacks

A skill library for people that make stuff

A Claude Code plugin bundling reusable skills for builders.

Install

# In Claude Code
/plugin marketplace add ragnar-pwninskjold/tech-snacks
/plugin install tech-snacks@tech-snacks

Or clone locally and point Claude Code at the directory.

Structure

tech-snacks/
├── .claude-plugin/
│   └── marketplace.json              # Marketplace catalog (for /plugin marketplace add)
├── .github/
│   └── PULL_REQUEST_TEMPLATE.md
├── plugins/
│   └── tech-snacks/                  # The plugin itself
│       ├── .claude-plugin/
│       │   └── plugin.json           # Plugin manifest (version lives here)
│       ├── agents/                   # Plugin agents (dispatched via Task)
│       │   └── research/             # caveman-search's research agents
│       │       ├── caveman-repo-scout.md
│       │       ├── caveman-verifier.md
│       │       └── caveman-web-researcher.md
│       ├── workflows/                # Multi-agent workflow scripts (.workflow.js)
│       │   ├── mine-claude-md-from-sessions.workflow.js
│       │   └── react-refactor-tournament.workflow.js
│       └── skills/                   # All skills live here as peer directories
│           ├── ui-cloner/
│           │   ├── SKILL.md
│           │   ├── references/       # Phase-by-phase procedures
│           │   ├── templates/        # Canonical output artifact shapes
│           │   └── examples/         # Worked examples for high-fidelity blocks
│           ├── prd-to-ux/
│           │   ├── SKILL.md
│           │   ├── references/
│           │   └── templates/
│           ├── lite-prd/
│           │   ├── SKILL.md
│           │   └── references/       # Traversal protocol + PRD output template
│           ├── scaffold-claude/
│           │   ├── SKILL.md
│           │   ├── references/
│           │   └── templates/
│           ├── intent-layer/         # Vendored (see Credits)
│           │   ├── SKILL.md
│           │   ├── LICENSE
│           │   ├── references/
│           │   └── scripts/
│           ├── caveman-search/
│           │   ├── SKILL.md
│           │   └── references/
│           ├── mine-claude-md/
│           │   └── SKILL.md          # Wraps the mine-claude-md-from-sessions workflow
│           └── react-refactor-tournament/
│               └── SKILL.md          # Wraps the react-refactor-tournament workflow
├── LICENSE
└── README.md

Skills

• ui-cloner — Site Replication Intelligence Protocol (SRIP). Full 4-phase pipeline for cloning a target site's UI: forensic audit → brand interview → synthesis → quality check, plus a post-build iterator for dialing in fidelity. Each phase is defined by an internal reference file under references/, with output shapes in templates/ and worked examples in examples/.
• prd-to-ux — translate a PRD or product idea into tool-agnostic screen prompts for web UX generators (Stitch, Figma AI, Pencil, Claude Design).
• lite-prd — turn a vague feature ask into a lightweight PRD by interviewing the user with AskUserQuestion, traversing a design tree (branching deeper when an answer opens new territory, --deep for exhaustive), persisting every Q&A round to docs/lite-prd/<slug>/qa-log.md, and emitting lite-prd.md.
• scaffold-claude — interview the user to scaffold a project-level CLAUDE.md / AGENTS.md, capturing edge cases and tribal knowledge.
• intent-layer — set up hierarchical AGENTS.md infrastructure so agents navigate codebases like senior engineers. Vendored from crafter-station/skills — see Credits.
• caveman-search — terse, research-backed answers to tech/tooling questions. Runs Compound-Engineering-style decomposed research (local repo + live web), verifies, and renders in a compressed "caveman" voice that leads with the load-bearing truth. Backed by three research agents in agents/research/.
• mine-claude-md — mine recent Claude Code sessions for non-obvious, multi-file CLAUDE.md candidates, adversarially verify them, and propose paste-ready additions. Wraps the mine-claude-md-from-sessions workflow.
• react-refactor-tournament — review React/Next.js code against the real vercel-react-best-practices skill, backlog the performance findings by rule id + impact tier, rank the most over-subscribed tiers, then fix + test the top N in isolated worktrees. Wraps the react-refactor-tournament workflow.

Adding a Skill