openclaw-qa-testing

Runs and debugs OpenClaw qa-lab and qa-channel end-to-end test scenarios.

Updated Jun 19, 2026
One-click install
npx skills add https://github.com/AmirulAndalib/Vilvona-AI --skill openclaw-qa-testing-amirulandalib
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: openclaw-qa-testing
Source: https://github.com/AmirulAndalib/Vilvona-AI/tree/main/.agents/skills/openclaw-qa-testing
Command: npx skills add https://github.com/AmirulAndalib/Vilvona-AI --skill openclaw-qa-testing-amirulandalib

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running OpenClaw's QA suite involves many moving parts: choosing mock versus live provider lanes, picking the right model policy, locating credentials in 1Password or Convex, and interpreting artifacts. This Skill encodes all of that operational knowledge so QA runs are consistent and reproducible. ## Core Features & Use Cases - Suite execution: Run the full qa-lab suite in mock or live-frontier mode with the correct model policy (openai/gpt-5.4, fast mode) and collect summary and report artifacts. - Specialized lanes: Execute Matrix live profiles, OTEL smoke tests, character/persona evals across multiple models, Codex CLI model lanes, and WhatsApp or Telegram live credential flows. - Credential handling: Look up QA secrets from 1Password or Convex leased credentials, with explicit rules for what to do when values are missing. - Use Case: A maintainer asks to validate a beta npm package against the live Telegram Docker lane; the Skill dispatches the GitHub workflow, polls artifacts, and reports pass/fail with artifact paths. ## Quick Start Ask the agent to run the OpenClaw QA suite in live-frontier mode and report the watch URL, pass/fail counts, and artifact paths.

Frequently Asked Questions about openclaw-qa-testing

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run the OpenClaw QA suite against a live OpenAI model?

Run pnpm openclaw qa suite with --provider-mode live-frontier and --model openai/gpt-5.4, passing OPENCLAW_LIVE_OPENAI_KEY from your OpenAI API key. Results are written to a summary JSON and markdown report under .artifacts/qa-e2e.

How do I run character or persona evals across multiple models?

Use pnpm openclaw qa character-eval with --model entries in the form provider/model,thinking=<level> and one or more --judge-model options. The report includes judge rankings, durations, and full transcripts for each candidate model.

Where are OpenClaw QA credentials stored?

QA secrets live in 1Password, primarily the OpenClaw vault item Telegram E2E, plus Convex items for site URL and maintainer or CI secrets. Use op only inside tmux, and ask the maintainer if a required value is missing rather than guessing.

Which OpenAI models are allowed for live QA lanes?

The live OpenAI lane uses openai/gpt-5.4 with fast mode enabled. The gpt-5.4-pro and gpt-5.4-mini variants are explicitly disallowed, and the model policy should only change when the user explicitly requests it.

Why does the live QA lane time out or fail to find repo files?

Short live timeouts are fixed by widening waits in extensions/qa-lab/src/suite.ts, and discovery failures are fixed by pointing prompts at repo/ paths inside the seeded workspace. A dirty tree can also trigger a pre-run build before ports appear.