Skip to content

Latest commit

 

History

History
92 lines (68 loc) · 3.71 KB

File metadata and controls

92 lines (68 loc) · 3.71 KB

QA Orchestra — Claude Code Instructions

@context/CONTEXT.md

What this project is

An open-source, multi-agent QA toolkit for Claude Code. Each QA task is handled by a specialized agent in .claude/agents/. Project context is auto-loaded from context/CONTEXT.md above.

Stack-agnostic. Works with any web application — React, Angular, Vue, Rails, Django, Spring Boot, or any other stack. All project-specific details live in context/CONTEXT.md.

Quick start

# 1. Clone into your workspace
git clone https://github.com/Anasss/qa-orchestra.git

# 2. Fill in your project context
cp examples/CONTEXT.example.md context/CONTEXT.md
# Edit context/CONTEXT.md with your stack details

# 3. Open Claude Code — agents auto-load from .claude/agents/
# Type @agent-name or describe what you need

How to invoke agents

# Full pipeline — checkout PR, run app, validate in browser, review diff, generate tests
"Run @orchestrator for PR #42 on my-backend and PR #38 on my-frontend"

# Direct invocation
"@environment-manager checkout feature/ISSUE-1 and start the app"
"@functional-reviewer compare this diff against these ACs: ..."
"@test-scenario-designer generate scenarios for these ACs: ..."
"@automation-writer read qa-output/test-scenarios.md and generate Playwright tests"
"@manual-validator guide me through qa-output/test-scenarios.md"
"@bug-reporter read qa-output/functional-review.md and create bug reports"
"@browser-validator validate scenarios against the running app"
"@release-analyzer analyze the diff between v1.0 and HEAD across all repos"
"@smart-test-selector which existing tests are affected by this diff?"

Agent map

Agent Model Reads Writes
orchestrator Sonnet ticket + inputs qa-output/plan.md
environment-manager Sonnet PR/branch + CONTEXT.md qa-output/environment-status.md
functional-reviewer Opus AC + diff + browser findings qa-output/functional-review.md
bug-reporter Sonnet functional-review + browser-validation qa-output/bug-reports.md
test-scenario-designer Sonnet AC qa-output/test-scenarios.md
automation-writer Sonnet test-scenarios.md qa-output/automation/
manual-validator Sonnet test-scenarios.md qa-output/validation-report.md
browser-validator Sonnet test-scenarios.md + Chrome MCP qa-output/browser-validation.md
release-analyzer Opus multi-repo diffs + AC qa-output/release-analysis.md
smart-test-selector Sonnet diff + existing tests qa-output/test-selection.md

Execution flow

1. environment-manager
   └── Checkout branches, install deps, start app, health check

2. (parallel)
   ├── functional-reviewer ──→ qa-output/functional-review.md
   └── test-scenario-designer ──→ qa-output/test-scenarios.md

3. browser-validator (Chrome MCP)
   └── Navigate app, execute scenarios, verify ──→ qa-output/browser-validation.md

4. (conditional, parallel)
   ├── bug-reporter (if gaps found) ──→ qa-output/bug-reports.md
   └── automation-writer ──→ qa-output/automation/

Output chaining

Every agent saves output to qa-output/. The next agent reads from that file. Never paste raw output between agents — always reference the file path.

Key principle: Live validation

The workflow is designed around testing running code, not just reading diffs. The environment-manager checks out the PR branch and runs the application locally. Browser validation happens against the actual running feature, not theoretical analysis.

Verification

Before running any agent, define what "done" looks like. If output looks wrong, stop — do not continue the chain with bad input.