Earn 14 free days when your bug report or suggestion is accepted — how it works

Tier-B runbook (Windows)

Tier-B sweeps validate CodeLoop on real Windows hosts for the Test 24 rows that can't be exercised from CI runners. This page is the public version of docs/TIER_B_RUNBOOK.md in the repo and pairs two PowerShell scripts: scripts/windows/preflight.ps1 (one-time host setup) and scripts/windows/capture-evidence.ps1 (per-row recording wrapper).

One-time host preflight

pwsh -File scripts\windows\preflight.ps1

Run as Administrator (chocolatey installs need it). The script installs node, git, ffmpeg,adb, flutter, dotnet (skipping anything already on PATH), checks that Cursor and Claude Code are installed, then runs npx codeloop auth, npx codeloop init --global, and npx codeloop doctor. The summary at the end is green / red so you can screenshot it as evidence. The script is idempotent — re-run it after a fresh checkout or after upgrading any IDE.

Per-row capture flow

Each Test 24 row uses the same wrapper but with a row-specific switch. Output lands intier-b-evidence/<row-slug>-<timestamp>/ and includes recording.mp4, verify.log, metadata.json, and (best-effort) dev-report.md.

Row 24.6 — Flutter Windows desktop

pwsh -File scripts\windows\capture-evidence.ps1 -RowId 24.6 -DurationSec 90

Row 24.7 — .NET / WPF / WinUI

pwsh -File scripts\windows\capture-evidence.ps1 -RowId 24.7 -DurationSec 90

Row 24.8 — Android Studio (native + Flutter mobile)

pwsh -File scripts\windows\capture-evidence.ps1 -RowId 24.8 -DurationSec 120

Row 24.9 — Web app via Playwright

pwsh -File scripts\windows\capture-evidence.ps1 -RowId 24.9 -DurationSec 60 -Url "https://staging.example.com"

Row 24.12 — Multi-monitor scenario

pwsh -File scripts\windows\capture-evidence.ps1 -RowId 24.12 -DurationSec 120 -Monitor "desktop"

Evidence appears in codeloop doctor

After any capture, npx codeloop doctorprints a “Tier-B / multi-app evidence” subsection enumerating which Test 24 rows have folders under tier-b-evidence/ on this host. Missing rows are listed with the exact remediation command. This is informational only — Tier-A is what gates PRs; Tier-B captures stay user-executed by design.

Mapping back to the master checklist

Each successful capture lets you flip the corresponding row in the E2E test checklist (rows 24.6, 24.7, 24.8, 24.9, 24.12). For the full per-row pass criteria and troubleshooting guide, see docs/TIER_B_RUNBOOK.md in the repo.