Summary

Windows source installs can misclassify the Control UI build result because scripts/install.ps1 tests the native command's output as a PowerShell Boolean instead of checking its process exit code.

Real environment observed

• Current openclaw/openclaw main
• Windows PowerShell native-command semantics
• -InstallMethod git source installation path

User-visible symptom

If pnpm ui:build fails after writing output, the installer can omit the existing failure warning and continue as though the UI build succeeded. Conversely, a successful command that produces no output can be treated as a failure.

Source-level reproduction on current main

The source install path currently uses:

if (-not (& $pnpmCommand ui:build)) {
    Write-Host "[!] UI build failed; continuing (CLI may still work)"
}

A controlled PowerShell probe reproduces the false negative:

{"LegacyOutputTruthinessWarns":false,"ExitCodeCheckWarns":true,"ExitCode":1}

The simulated native build emits output and sets exit code 1. Output truthiness suppresses the warning, while checking $LASTEXITCODE detects the failure.

Expected behavior

The installer should classify pnpm ui:build from its native process exit code, independent of whether the command writes output.

Actual behavior

The command's output is converted to a Boolean, so output volume can change the reported build status.

Scope / non-goals

This report is limited to source-install UI build status detection. The UI build remains non-fatal, and no npm, pnpm, or runtime behavior needs to change.

Proposed fix

Run the build normally, then inspect $LASTEXITCODE. Add a regression assertion that rejects the output-truthiness pattern.

Live searches for the affected command, installer warning, and exit-code behavior found no matching open or closed issue or PR before filing.