azanir QA Orchestrator
Purpose
Master coordinator for all QA skills in the ecosystem. Routes work through pipeline modes, formalized handoff chains, and scheduler rules. Selects frameworks and languages based on project context. Uses Memory MCP for cross-session persistence.
Trigger Phrases
- "Run full QA cycle" / "Full-cycle pipeline"
- "Docs only" / "Generate QA docs"
- "Test cases only" / "Testcase pipeline"
- "Write tests" / "Test writing pipeline"
- "Report only" / "Generate test report"
- "Orchestrate [chain name]" / "Run Chain 1" / "E2E chain" / "API-first chain"
- "After merge" / "Before release" / "Requirement changed"
- "Select framework for [project]"
Pipeline Modes
| Mode | Flow | When to Use |
|---|---|---|
| full-cycle | docs → test cases → tests → execution → reports | New feature, full validation |
| docs-only | requirements → spec → diagram | Documentation phase |
| testcases-only | docs/spec → test cases | Test design phase |
| write-tests | test cases → tests | Automation phase |
| report | results → report | Post-execution reporting |
See references/pipeline-modes.md for detailed configurations.
Formalized Handoff Chains
Chain 0 — Discovery → Full E2E
discovery-interview → requirements-generator → spec-writer → diagram-generator → testcase-from-docs → playwright-ts-writer → test-healer (if fail) → test-reviewer → test-reporter → project-memory:update
Chain 1 — Full E2E
requirements-generator → spec-writer → diagram-generator → testcase-from-docs → playwright-ts-writer → test-healer (if fail) → test-reviewer → test-reporter → project-memory:update
Chain 2 — UI-first
browser-data-collector → testcase-from-ui → playwright-ts-writer → test-healer → coverage-analyzer → task-creator (for gaps) → project-memory:update
Chain 3 — API-first
api-contract-curator → testcase-from-docs → supertest-writer / httpx-writer → pact-writer → test-reporter → project-memory:update
Chain 4 — Stabilization
flaky-detector → test-healer → test-reviewer → changelog-analyzer → regression scope → project-memory:update
See references/handoff-chains.md for input/output contracts per step.
Scheduler Rules
| Trigger | Actions |
|---|---|
| After merge | Smoke/regression → report → bug if fail |
| Before release | Completion report + risk analysis |
| After requirement change | Impact analysis → update RTM |
See references/scheduler-rules.md for automation patterns.
Framework / Language Selection
| Project Signal | Ecosystem | Writers |
|---|---|---|
tsconfig.json present |
TypeScript | playwright-ts-writer, jest-writer, vitest-writer, supertest-writer, cypress-writer |
setup.py / pyproject.toml |
Python | playwright-py-writer, pytest-writer, httpx-writer, selenium-py-writer |
pom.xml / build.gradle |
Java | selenium-java-writer, junit5-writer, rest-assured-writer, spring-test-writer |
Selection is automatic based on project root detection. User can override.
Memory MCP
Uses Memory MCP for cross-session persistence:
- Last pipeline mode and chain used
- Project context (framework, language)
- Scheduler state (last run, next trigger)
- Handoff artifacts (IDs, paths) for resume
Workflow
- Input — User request (pipeline mode, chain, or scheduler trigger)
- Context — Detect project (tsconfig, pyproject, pom.xml)
- Route — Select chain or pipeline per mode
- Execute — Invoke skills in sequence; pass outputs as inputs
- Persist — Store state via Memory MCP
- Record — Auto-update
docs/qa-memory/viaqa-project-memory(final chain step) - Output — Final artifact (report, tests, docs) + summary
Scope
Can do (autonomous):
- Route to any pipeline mode or handoff chain
- Select framework/language from project structure
- Invoke skills in sequence with handoff contracts
- Persist state via Memory MCP
- Apply scheduler rules when triggered
Cannot do (requires confirmation):
- Change project scope or release criteria
- Override organizational test policy
- Execute tests in production
Will not do (out of scope):
- Modify production code or environments
- Approve releases (stakeholder responsibility)
- Create skills not in ecosystem
Quality Checklist
- Pipeline mode matches user intent
- Chain selected aligns with project type (E2E, UI, API, stabilization)
- Framework/language detected correctly
- Each handoff passes valid input per contract
- Scheduler rules applied when trigger matches
- State persisted for resume
- Final output (report, tests, docs) produced
Troubleshooting
| Symptom | Likely Cause | Fix |
|---|---|---|
| Wrong chain selected | Ambiguous project type | Ask: E2E, UI-first, or API-first? |
| Missing handoff | Skill output format mismatch | Check references/handoff-chains.md contracts |
| Framework not detected | Non-standard project layout | User override: specify ts/py/java |
| Scheduler not firing | Trigger not matched | Verify "after merge" / "before release" phrasing |
| State lost between sessions | Memory MCP not configured | Ensure MCP server enabled and connected |