Goal Suite Patterns
Use stable tags so reruns can be compared across machines and dates:
[suite:<name>] [baseline:<git-sha-or-date>] [matrix:<targets>] [exit:<definition-of-done>]Recommended recurring families:
openclaw-qa-all-surfaces: RPC, kitchen sink, onboarding, plugins, SDKs, QA Lab, and scripts.openclaw-cross-platform: macOS ARM64, Windows native, WSL2, Linux, and Docker.local-model-code-mode: Ollama, Hugging Face, constrained models, and a frontier control.tool-schema-safety: malformed schemas, dynamic tools, doctor, UI, and channel failure behavior.sandbox-compatibility: OpenShell, restricted filesystem/network, browser, packages, and subprocesses.localization-all-surfaces: Control UI, web, docs UI, iOS, Android, macOS, and watch.dead-code-complexity: core, apps, Control UI, plugins, and SDKs with no contract change.contributor-pr-batch: bounded low-risk PR queue work; keep separate from product beta suites.
Every rerun should record:
- exact baseline SHA or release;
- unchanged test matrix;
- product failures versus infrastructure blockers;
- before/after reliability or performance evidence;
- landed PRs and remaining issues;
- skipped surfaces and the next rerun trigger.