Unit and focused integration coverage for OpenClaw.Shared.
Do not maintain per-file test counts in this README. They drift whenever
theories or focused regressions are added. The repository-wide source inventory
and dated required-suite runtime snapshots live in
docs/TEST_COVERAGE.md.
Use the current run's report for completion claims. For new regressions, follow
test authoring.
From the repository root:
$env:OPENCLAW_REPO_ROOT = (Get-Location).Path
dotnet test .\tests\OpenClaw.Shared.Tests\OpenClaw.Shared.Tests.csprojAfter the project has already restored and built:
dotnet test .\tests\OpenClaw.Shared.Tests\OpenClaw.Shared.Tests.csproj --no-restoreRun a focused class or method:
dotnet test .\tests\OpenClaw.Shared.Tests\OpenClaw.Shared.Tests.csproj `
--filter "FullyQualifiedName~ExecApprovalsCoordinatorTests"- Gateway WebSocket protocol parsing and request construction
- Session, channel, usage, node, command, and display models
- Device identity, token storage, signing, and migration
- Node capability contracts for system, canvas, screen, camera, location, speech, device, browser proxy, and app commands
- Local MCP authentication, JSON-RPC dispatch, cancellation, and telemetry
- Windows V2 exec approval validation, normalization, evaluation, prompting, persistence, and execution-boundary revalidation
- MXC availability, config/policy construction, command routing, and focused integration behavior
- URL, canvas, web-bridge, notification, and secret-handling security helpers
- Audio model download integrity and speech-language helpers
Some integration tests require explicit environment gates or local Windows
capabilities and may skip on unsupported hosts. A successful zero-test
--no-restore run in a fresh worktree is not proof; build the project first or
omit --no-restore.