Supercharge Claude Code with skills that make test automation, debugging, and AI-assisted QA faster, smarter, and tighter.
One script to install all essential Claude Code plugins for QA/SDET work. Run these install commands from within Claude Code.
#!/usr/bin/env bash
# ─────────────────────────────────────────────────────────
# qaskills.sh — Claude Code Plugin Installer for QA/SDET
# Run each line inside Claude Code (not in your shell)
# ─────────────────────────────────────────────────────────
# 1. Superpowers — full dev methodology (TDD, agents, plans)
/plugin install superpowers@claude-plugins-official
# 2. Caveman — ~75% token savings, terse mode for daily use
/plugin install caveman@caveman
# 3. Frontend Design — bold, distinctive UI generation
/plugin install frontend-design@claude-plugins-official
# 4. Everything Claude Code — coding standards + security + TDD patterns
/plugin install everything-claude-code@everything-claude-code
# 5. MCP Server Dev — build MCP tools that connect to any API
/plugin install mcp-server-dev@claude-plugins-official
# 6. Skill Creator — write your own reusable skills
/plugin install skill-creator@claude-plugins-official
# 7. Playwright — official Playwright testing assistant
/plugin install playwright@claude-plugins-officialAdd --scope user to install globally across all projects, or --scope project to pin to one repo. Default is user scope. Use /plugin list to verify installed plugins.
A complete software development methodology baked into Claude Code. Superpowers intercepts every task — exploring intent, writing specs, building plans, executing via subagents, and reviewing output. The most impactful single install.
/plugin install superpowers@claude-plugins-officialTriggers before ANY creative work. Explores user intent and requirements before a single line of code is written.
auto — creative tasksWhen you have a spec, converts it to a step-by-step implementation plan clear enough for any subagent to follow.
auto — before multi-step tasksEnforces true red/green TDD. Write test first, fail it, implement, pass. No exceptions.
auto — any feature/bugfixStructured root-cause analysis before any fix is proposed. No shotgun debugging.
auto — any bug or test failureDispatches parallel subagents to execute independent tasks from a plan. Hours of autonomous work without deviation.
/subagent-driven-developmentWhen 2+ tasks are independent — spawns concurrent agents, merges results. Cuts wall-clock time dramatically.
auto — 2+ independent tasksForces verification commands to run before any "done" claim. Evidence before assertions — always.
auto — before any PR/commitIsolates feature work in a clean git worktree before executing plans. No cross-contamination of in-progress work.
auto — before plan executionStructured code review request. Ensures the reviewer gets all context: what changed, why, what to watch for.
auto — after task completionBefore implementing review feedback — rigorously checks if feedback is technically correct. No blind agreement.
auto — when feedback arrivesWhen implementation is done and tests pass — guides merge, PR creation, or cleanup decision.
auto — end of feature branchCreate your own reusable skills: structure, triggers, content format, deployment and verification.
/writing-skillsThe TDD + systematic-debugging + verification-before-completion trio prevents the #1 QA failure mode: tests that pass locally but lie. Superpowers enforces discipline at the skill level, not the prompt level.
Caveman mode cuts Claude's output tokens by ~75% while keeping every technical detail. Same fix, 75% fewer words. Brain still big. Three intensity levels, terse commits, compressed reviews, and a cavecrew of specialized subagents.
/plugin install caveman@caveman"The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. When you pass an inline object as a prop, React's shallow comparison sees it as a different object every time..."
"New object ref each render. Inline object prop = new ref = re-render. Wrap in useMemo."
Core mode. Three levels: lite (drop filler), full (fragments OK), ultra (minimal grunt). Auto-switches off for security warnings and irreversible ops.
/caveman · /caveman lite · /caveman ultraCompress CLAUDE.md, todos, preferences into caveman format. Cuts ~46% of input tokens per session on memory files.
/compressUltra-terse commit messages. Conventional Commits format. Subject under 50 chars. Body only when "why" is non-obvious.
/caveman-commitUltra-compressed code review. One line per finding: location, problem, fix. No praise. No scope creep. Severity-tagged.
/caveman-reviewShows real token usage and estimated savings for the current session. Reads directly from the Claude Code session log.
/caveman-statsQuick-reference card for all caveman modes, skills, and commands. One-shot display, not a persistent mode.
/caveman-helpDecision guide for delegating to 3 specialized subagents: investigator (locate code), builder (surgical edits), reviewer (diff review).
/cavecrewcavecrew-investigator — read-only code locator, great for "where is X defined". cavecrew-builder — surgical 1-2 file edits, perfect for typo fixes and single-function rewrites. cavecrew-reviewer — diff/branch reviewer with severity-tagged findings.
| Level | Command | What changes | Token savings |
|---|---|---|---|
| lite | /caveman lite | Drop filler words and pleasantries only | ~30% |
| full (default) | /caveman | Drop articles, fragments OK, short synonyms | ~75% |
| ultra | /caveman ultra | Minimal output, grunt-level compression | ~85% |
One skill that makes Claude produce production-grade UIs instead of generic AI-looking interfaces. Bold aesthetic choices, distinctive typography, high-impact animations. Critical for practice pages and dashboards.
/plugin install frontend-design@claude-plugins-officialGenerates distinctive, production-grade frontend code. Triggers on "create a dashboard", "build a landing page", "design a settings panel". Avoids generic AI aesthetics.
auto — any UI taskWhen building TTA practice pages (forms, tables, dialogs), frontend-design ensures each page has a distinct visual identity instead of cookie-cutter Bootstrap layouts. Students recognize the target elements faster.
# These prompts auto-trigger the frontend-design skill:
"Create a QA dashboard for test run metrics"
"Build a login form practice page with dark mode"
"Design a settings panel for a test runner config"
"Create a table with sortable columns and pagination"The complete collection of Claude Code configs from an Anthropic hackathon winner — 10+ months of intensive daily use building real products, distilled into production-ready skills for coding standards, security, TDD, and continuous learning.
/plugin install everything-claude-code@everything-claude-codeUniversal coding standards for TypeScript, JavaScript, React, and Node.js. Applied automatically to new code.
auto — any code taskComprehensive security checklist: auth flows, user input, secrets, API endpoints, payment handling. Auto-triggers on sensitive code.
auto — auth / API / sensitive codeEnforces 80%+ test coverage with unit, integration, and E2E tests. Write tests first — no shortcuts.
auto — new features / bug fixesAPI design, database optimization, server-side best practices for Node.js, Express, and Next.js API routes.
auto — backend tasksReact, Next.js, state management, performance optimization, and UI best practices. Complements frontend-design.
auto — React / Next.js tasksAutomatically extracts reusable patterns from each Claude Code session and saves them as learned skills for future use.
auto — end of sessionsCheckpoint vs continuous evals, grader types, pass@k metrics. Prevents premature "done" claims.
auto — before completionsSuggests manual context compaction at logical intervals. Preserves context through task phases rather than arbitrary auto-compaction.
auto — long sessionsThree skills for designing and building MCP (Model Context Protocol) servers that connect Claude to any API, tool, or data source. Perfect for building QA-specific MCP tools — Jira, TestRail, Allure, CI pipelines.
/plugin install mcp-server-dev@claude-plugins-officialEntry point. Interrogates use case, picks deployment model, picks tool-design pattern, routes to specialized skill. Start here.
/mcp-server-dev:build-mcp-serverAdds interactive UI widgets (forms, pickers, confirm dialogs) rendered inline in Claude chat. For MCP servers with rich interaction.
/mcp-server-dev:build-mcp-appPackages a local stdio server with its runtime so users install it without Node/Python. For servers that need local machine access.
/mcp-server-dev:build-mcpbBuild an MCP tool that connects Claude to Jira (auto-generate test plans from tickets), TestRail (sync results), or your CI pipeline (trigger runs, fetch logs). The CrewAI + Jira MCP tutorial on this site uses exactly this pattern.
Write your own reusable skills and deploy them across projects. If you have a repeating QA workflow — practice page builder, test case generator, API contract checker — turn it into a skill once and invoke it forever.
/plugin install skill-creator@claude-plugins-official# Ask Claude to create a skill for your QA workflow:
"Create a skill called 'api-contract-check' that runs whenever
I say 'check this API contract'. It should verify that the
OpenAPI spec matches the actual endpoint responses using
Playwright's request interception."
# Claude will use skill-creator to structure it properly,
# including triggers, content, and deployment instructions.Official Anthropic Playwright plugin — brings deep Playwright knowledge, test patterns, fixture design, and debugging guidance directly into Claude Code. Essential for the entire TTA Playwright curriculum.
/plugin install playwright@claude-plugins-officialInstall the Playwright plugin, then open any page from the TTA Playwright practice library. Claude will know exactly which locators to use, how to write fixtures, and how to set up CI for the exercises you're practicing.
Quick reference — all 7 plugins, install scope, and primary use case for QA/SDET work.
| Plugin | Skills | Source | Best for | Auto-trigger? |
|---|---|---|---|---|
| superpowers | 14 | Official | Full dev methodology — TDD, debugging, parallel agents | ✓ Yes |
| caveman | 7 + 3 agents | Community | Token savings, terse commits, compressed reviews | ✓ After /caveman |
| frontend-design | 1 | Official | Production-grade, distinctive UI generation | ✓ Yes |
| everything-claude-code | 8+ | Community | Coding standards, security, TDD, continuous learning | ✓ Yes |
| mcp-server-dev | 3 | Official | Building MCP tools for Jira, CI, TestRail | Manual |
| skill-creator | 1 | Official | Writing your own reusable QA skills | Manual |
| playwright | TBD | Official | Playwright test patterns, fixtures, CI config | ✓ Yes |
Superpowers and everything-claude-code both include TDD enforcement. They complement rather than conflict, but you may see double prompts for TDD on some tasks. Start with superpowers alone — add everything-claude-code only if you want the extra coding-standards and security-review auto-triggers.
# Check what's installed:
/plugin list
# See all available plugins in the official marketplace:
/plugin search
# Update a specific plugin:
/plugin update superpowers@claude-plugins-official
# Remove a plugin:
/plugin remove caveman@caveman