testland-qa

A rigorously curated quality-engineering plugin marketplace for Claude Code;
its skills also install into any agent via the skills.sh CLI. Every component
is reviewed before merge.

Why testland-qa

• Six-dimension quality rubric applied at manual PR review, with a hard
  reject for uncited claims (citation theater) as the citation floor
• CI-validated composition: every agent's preloaded skills are
  reference-checked, no dangling deps
• Differentiation required: every component must articulate how it differs
  from its nearest neighbors. Generic, persona-shaped scopes that can't name a
  trigger condition get sent back for reshaping
• Reviewer-calibrated: two-evaluator rubric, A/C/F-grade exemplars in
  docs/REVIEWER_TRAINING.md

See Quality bar and docs/REVIEWER_CHECKLIST.md.

How it works

The marketplace ships these building blocks:

• Role bundle: a one-command install of a whole role. qa-starter (the
  essentials every tester needs) plus the qa-role-* packs. Installing a bundle
  installs its member plugins for you. Start here.
• Plugin: an installable unit scoped to one QA area (e.g. qa-api-testing,
  qa-load-testing). Install only the plugins your stack needs.
• Skill: an atomic, self-contained capability inside a plugin, usually
  wrapping one tool or one technique (e.g. great-expectations,
  oauth-flow-test-author). Claude loads a skill when your request matches its
  trigger; you can also ask for it by name.
• Agent: a task-scoped subagent that runs one focused job (e.g.
  schema-diff-reviewer reviews a migration diff and returns a findings table).
  An agent may preload one or more skills to do its work.

Installed components stay dormant until a matching task comes up, so adding a
plugin doesn't add noise. It adds capability that activates on demand.

Install

Install a whole role in one command (recommended)

Add the marketplace once, then install a role bundle. Claude Code pulls in all of
its member plugins automatically:

/plugin marketplace add testland/qa
/plugin install qa-starter@testland-qa

qa-starter is the essentials every tester needs. For a stack-specific role,
install one of the qa-role-* bundles instead:

/plugin install qa-role-security@testland-qa    # AppSec / security testing
/plugin install qa-role-frontend@testland-qa    # web / UI automation
/plugin install qa-role-backend@testland-qa     # API / microservices / distributed

One bundle install brings in its whole member set (Claude Code lists what it
added). The role bundles are listed under Start here and in
CATALOG.md. Requires Claude Code v2.1.110+ (v2.1.143+ to enable
the set together).

Install individual plugins

Prefer a narrower set? Install plugins one at a time:

/plugin install qa-data-quality@testland-qa

Cross-agent install (skills.sh)

Not on Claude Code? Any skill installs into Cursor, Copilot, Gemini, Windsurf,
and 20+ other agents via the skills.sh CLI:

npx skills add testland/qa                              # browse and install from the repo
npx skills add testland/qa --skill <skill>              # one specific skill
npx skills add testland/qa --skill <one> --skill <two>  # several at once

The CLI installs skills only. Agents and the role bundles (qa-starter and
the qa-role-* family) need Claude Code: a bundle owns no SKILL.md of its
own, so npx skills add on one reports "No skills found".

Direct URL

/plugin marketplace add https://github.com/testland/qa

Manual / hermetic environments

git clone https://github.com/testland/qa ~/.claude/marketplaces/testland-qa

Before you install: plugins run inside your Claude Code session and ship
agent instructions and tool wrappers. Anthropic doesn't vet marketplace
contents, so review a plugin's components before installing it into a sensitive
project. Every component here is reviewed before merge (see
Quality bar), but you remain in control of what runs.

Start here

New to the marketplace? Two ways in, depending on how much you want at once.

Try one or two first

Install a couple of plugins for your role rather than everything. Components
activate on demand, so a focused set keeps things sharp.

Or adopt a whole role at once

A role bundle installs a curated set for your role in a single command.
/plugin install <bundle>@testland-qa pulls in every member plugin. Start with
qa-starter, or pick a role:

Bundles overlap where roles do (installing a plugin twice is harmless). The
per-language unit-test plugins and a few language-agnostic extras stay à la
carte; add them on top of a bundle as your stack needs.

Using an installed plugin

Once a plugin is installed, its skills and agents are available to Claude Code.
Invoke them by describing the task in plain language. Example with
qa-data-quality:

/plugin install qa-data-quality@testland-qa

• Ask "add Great Expectations checks to my orders pipeline" and the
  great-expectations skill scaffolds an ExpectationSuite + Checkpoint and wires
  the results into a CI gate.
• On a database change, ask "review this migration's schema diff" and the
  schema-diff-reviewer agent returns a Critical / Warning / Info findings table
  covering breaking-vs-additive changes and downstream impact.

Each plugin's README.md lists its skills and agents and what each one does.

Plugin catalog

Plugins are grouped into ten categories. For the full, always-current list with
per-plugin links, versions, and component counts, see CATALOG.md.
New here? Start with the curated core tier in CORE.md — the
highest-demand skills and agents across the catalog.

• Role bundles: one-command role installs, each bundling a curated capability set
• Foundations: test process, environment, data, reporting, impact analysis, test review, test management, team management
• Functional testing: API, BDD, E2E, mobile, desktop, embedded, game, contract, mutation, property-based, per-language unit tests
• Quality engineering: data quality, visual regression, accessibility, localization, charts, PDF/print, modern web, browser extension, PWA
• Security and compliance: SAST, DAST, SCA, secrets, SBOM, fuzzing, compliance, multi-tenancy isolation, test-data privacy
• Operations and resilience: flake triage, bug repro, defect management, chaos, resilience drills, shift-left/right, load
• Backend and distributed systems: DB migrations, async jobs, caching, concurrency, distributed tracing, saga/CQRS, serverless, time and timezones
• Integrations and protocols: GraphQL, gRPC, real-time protocols, auth flows, notifications, payment, feature flags, experimentation
• AI and ML: LLM evaluation, ML model testing, AI-assisted test generation, data notebooks, search relevance
• Tooling: IaC, CI integration, CLI tools, code quality, compatibility, manual testing

Quality bar

Every component is reviewed against six dimensions before merge (the D1-D6
rubric). This is a manual PR review: a reviewer applies the rubric to the
diff via the D1-D6 checklist in .github/pull_request_template.md.

The merge bar is reviewer judgment: each dimension should clear its anchor, with
citations (D6) as the hard floor. Uncited "citation theater" is the dominant
failure mode and a hard reject. There is no automated rating gate and no
rating/d6 frontmatter; the rubric lives in the reviewer's checklist.
Mechanical lint (naming, JSON, description/body length, Windows-path hygiene)
still runs in CI via npm run validate and npm run audit.

See docs/REVIEWER_CHECKLIST.md for the rubric and
docs/CONTRIBUTING.md for the framework details.

Repository layout

testland-qa/
  .claude-plugin/marketplace.json    # required; lists all plugins
  plugins/                           # one folder per plugin
  templates/                         # plugin / skill / agent / command scaffolds
  scripts/                           # validate, scaffolding,
                                     # composition graph, inventory, catalog
  .github/                           # workflows, issue + PR templates,
                                     # CODE_OF_CONDUCT, SECURITY, CODEOWNERS
  docs/                              # CONTRIBUTING, PLUGIN_AUTHORING,
                                     # REVIEWER_CHECKLIST, REVIEWER_TRAINING,
                                     # COMPOSITION
  CLAUDE.md                          # Claude-native contribution guide
  CHANGELOG.md                       # release notes
  CATALOG.md                         # auto-generated plugin catalog
  Makefile                           # validate / compose / catalog targets
  LICENSE                            # MIT

Questions and feedback

• Bug or a broken component? Open an issue.
• Want a plugin that doesn't exist yet? Use the plugin-request issue template.
• General questions or ideas? Start a thread in Discussions.
• Security report? See SECURITY.md; don't open a public issue.

Contributing

Read CLAUDE.md and docs/CONTRIBUTING.md
before opening a PR. Use the scaffolder:

npm run new-plugin -- <plugin-name> "<one-line-description>" <primary-keyword>

Run validation locally before pushing:

make all   # validate + compose + catalog

Or, without make:

npm run validate
npm run compose

License

MIT. See LICENSE.