QA Boot

Bootstrap QA context for AI coding agents.

QA Boot scans a repository or multi-repo workspace, collects deterministic testing/build/CI evidence, captures unknowns, and generates Claude Code-compatible QA context files.

It helps QA engineers and quality leads onboard into messy projects without letting AI agents invent missing process knowledge.

QA Boot is not a test generator or QA automation framework. It is the context and safety layer before AI-assisted QA work.

Core idea

Most AI QA tooling focuses on generating, executing, or reviewing tests.

QA Boot focuses on the missing layer before that:

• What does this project know?
• What is unknown?
• Which facts are observed, inferred, or human-provided?
• Which quality risks are technical versus business-critical?
• What is Claude allowed to assume?
• What should Claude ask before acting?

V1 product boundary

V1 is a CLI-first file generator.

It should:

• scan repos deterministically,
• consume evidence providers like QA Radar,
• generate qa-context/,
• generate Claude Code project skills,
• generate unknowns and maturity summaries,
• avoid external writes by default.

V1 should not:

• auto-comment on PRs,
• auto-create Jira tickets,
• update TestRail,
• trigger CI,
• approve releases,
• access secrets,
• act as a full autonomous QA agent.

Install & usage

Requires Node 20+. Run qa-boot directly in any repo — no install step needed:

npx qa-boot init --project-name my-service
npx qa-boot scan        # auto-runs generate

scan detects languages/tests/CI/docs/build/agent-config, writes qa-context/facts.json, and (unless --no-generate) renders the qa-context/*.md summaries, unknowns.md, maturity.md, CLAUDE.qa.md, and the .claude/skills.

Record human-told QA knowledge (answers to open unknowns, or free-form facts):

npx qa-boot tell <unknown-id> "the answer" --by alice
npx qa-boot tell "we release every Tuesday" --domain release --by alice

Prefer a permanent command? npm install -g qa-boot gives you qa-boot on your PATH; npm install -D qa-boot pins a version per-project.

QA Radar is consumed live when installed:

npx qa-boot scan --with-qaradar   # spawns `qaradar analyze --json-output`
npx qa-boot scan --skip-qaradar   # built-in scanners only

When qaradar is absent the scan still succeeds and records the gap as an unknown (repo-risk.md is only written when QA Radar ran).

Developing qa-boot itself

git clone https://github.com/MuratKus/qa-boot.git
cd qa-boot && npm install
npm test                 # vitest
npm run dev -- scan      # run the CLI from source via tsx

Design and plan: docs/superpowers/specs/2026-06-02-qa-boot-v0-design.md and docs/superpowers/plans/2026-06-02-qa-boot-v0.md.

Start here

Read SPEC_INDEX.md first.

For implementation planning, start with:

• docs/00-overview.md
• docs/02-v1-scope.md
• docs/03-architecture.md

Tagline

Make Claude more grounded before making it more autonomous.