Copilot agent infrastructure, emulator reliability, and try-fix workflow improvements

[Copilot]

Improvements to Copilot agent infrastructure scripts, emulator reliability, and try-fix workflow.

Agent infrastructure

• Update model references from claude-opus-4.5 → claude-opus-4.6 in PLAN-TEMPLATE.md and SHARED-RULES.md
• Add environment blocker stop rules and mandatory cleanup between try-fix attempts in post-gate.md
• Prevent git add -f of state files in CustomAgentLogsTmp/ to avoid data loss on tree restore (try-fix/SKILL.md)

Review-PR.ps1 hardening

• Add -LogFile parameter using Start-Transcript (replaces external tee pipe that caused hangs)
• Restore working tree to clean state between phases to prevent skill file corruption
• Run post-ai-summary-comment.ps1 directly instead of via Copilot CLI
• Write finalize output to separate file to prevent main state file overwrite

Emulator/simulator reliability

• Rewrite Start-Emulator.ps1 with proper Android SDK discovery, AVD boot detection, and package manager readiness checks
• Add iOS simulator architecture detection (arm64 vs x64) in Build-AndDeploy.ps1 with fallback
• Increase Android emulator boot timeout from 2 → 10 minutes in android.cake

Utility

• Add helper functions in shared-utils.ps1

Original prompt

Create a pr with all the changes from this branch except the yml ci changes
https://github.com/dotnet/maui/tree/copilot-ci

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[PureWeen]

@copilot remove all the agent-pr-session files from this pr

[Copilot]

@copilot remove all the agent-pr-session files from this pr

Done in e2324ee — restored all 5 agent-pr-session files so they're no longer part of this PR.

[PureWeen]

@copilot update the title and description to match what's actually in this pr

The branch this was pulled from isn't relevant

[Copilot]

@copilot update the title and description to match what's actually in this pr

The branch this was pulled from isn't relevant

Updated the PR description to describe the actual changes: agent infrastructure updates, Review-PR.ps1 hardening, emulator/simulator reliability improvements, and utility additions. Removed all references to the source branch.

[Copilot]

Pull request overview

This PR updates the repo’s Copilot agent workflow docs/scripts and device-test infrastructure to improve try-fix iteration hygiene, PR review automation reliability, and emulator/simulator startup robustness.

Changes:

• Hardened PR review automation (Review-PR.ps1) with transcript logging, phase-to-phase workspace cleanup, and direct summary-comment posting.
• Reworked Android emulator + iOS simulator selection/boot logic and iOS simulator RID/app bundle selection.
• Updated try-fix / agent docs for 6-model sequencing and mandatory cleanup between attempts; increased Android emulator boot timeout.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
eng/devices/android.cake	Increases emulator boot timeout to reduce flaky boots.
.github/skills/try-fix/SKILL.md	Updates restore instructions to git checkout HEAD -- . and clarifies state file handling.
.github/scripts/shared/shared-utils.ps1	Adds a shared Write-Warning helper for consistent script output.
.github/scripts/shared/Start-Emulator.ps1	Reworks Android SDK/emulator discovery + boot readiness checks; broadens iOS simulator auto-selection logic.
.github/scripts/shared/Build-AndDeploy.ps1	Detects iOS simulator arch/RID and selects the correct built .app bundle with fallback.
.github/scripts/Review-PR.ps1	Adds -LogFile transcript support, restores clean state between phases, runs summary post script directly, and cleans orphaned copilot processes.
.github/agents/pr/post-gate.md	Updates try-fix sequencing to 6 models and mandates cleanup between attempts; expands blocker guidance and recovery notes.
.github/agents/pr/SHARED-RULES.md	Updates multi-model configuration to 6 models.
.github/agents/pr/PLAN-TEMPLATE.md	Updates plan template references for 6-model try-fix workflow.
.github/agents/pr.md	Updates Phase 4 multi-model configuration reference to 6 models.

[Copilot]

On Windows, $emulatorBin = Join-Path $androidSdkRoot "emulator/emulator" will usually not exist because the SDK ships emulator.exe. Since Test-Path checks the literal path (no PATHEXT resolution), this will cause the script to fail even when the emulator is installed. Consider using an OS-specific emulator path (e.g., emulator\emulator.exe on Windows) or resolving the emulator location via Get-Command emulator and/or checking both filenames.

Suggested change

	$emulatorBin = Join-Path $androidSdkRoot "emulator/emulator"
	# Check emulator binary exists
	if (-not (Test-Path $emulatorBin)) {
	Write-Error "Emulator binary not found at: $emulatorBin"
	$emulatorDir = Join-Path $androidSdkRoot "emulator"
	if ($IsWindows) {
	# On Windows the SDK ships emulator.exe; prefer that, then fall back to emulator without extension
	$emulatorBin = Join-Path $emulatorDir "emulator.exe"
	if (-not (Test-Path $emulatorBin)) {
	$emulatorBin = Join-Path $emulatorDir "emulator"
	}
	}
	else {
	# On macOS/Linux the binary is typically named 'emulator' without extension
	$emulatorBin = Join-Path $emulatorDir "emulator"
	}
	# Check emulator binary exists; if not, try resolving via PATH
	if (-not (Test-Path $emulatorBin)) {
	$emulatorCmd = Get-Command emulator -ErrorAction SilentlyContinue
	if ($emulatorCmd) {
	$emulatorBin = $emulatorCmd.Path
	}
	}
	if (-not (Test-Path $emulatorBin)) {
	Write-Error "Emulator binary not found. Last attempted path: $emulatorBin"

[Copilot]

The Android path unconditionally runs bash -c "pgrep ..." to verify the emulator process started. This will break on Windows environments that don't have bash/pgrep available (even though this script is used for Android on Windows). Use a Windows-compatible process check (e.g., Get-Process/WMI) or guard the bash calls behind if (-not $IsWindows).

Suggested change

	$emulatorProcs = bash -c "pgrep -f 'qemu.*$selectedAvd' || pgrep -f 'emulator.*$selectedAvd' || true" 2>&1
	if ([string]::IsNullOrWhiteSpace($emulatorProcs)) {
	Write-Error "Emulator process did not start. Checking log..."
	if (Test-Path $emulatorLog) {
	Get-Content $emulatorLog | Select-Object -Last 50 | ForEach-Object { Write-Info " $_" }
	}
	exit 1
	}
	Write-Info "Emulator process started (PIDs: $emulatorProcs)"
	if ($IsWindows) {
	# On Windows, use native process inspection instead of bash/pgrep
	$emulatorProcs = Get-Process | Where-Object {
	$_.ProcessName -match 'emulator' -or $_.ProcessName -match 'qemu'
	}
	if (-not $emulatorProcs) {
	Write-Error "Emulator process did not start. Checking log..."
	if (Test-Path $emulatorLog) {
	Get-Content $emulatorLog | Select-Object -Last 50 | ForEach-Object { Write-Info " $_" }
	}
	exit 1
	}
	$emulatorProcIds = $emulatorProcs | Select-Object -ExpandProperty Id -ErrorAction SilentlyContinue
	Write-Info "Emulator process started (PIDs: $($emulatorProcIds -join ', '))"
	}
	else {
	$emulatorProcs = bash -c "pgrep -f 'qemu.*$selectedAvd' || pgrep -f 'emulator.*$selectedAvd' || true" 2>&1
	if ([string]::IsNullOrWhiteSpace($emulatorProcs)) {
	Write-Error "Emulator process did not start. Checking log..."
	if (Test-Path $emulatorLog) {
	Get-Content $emulatorLog | Select-Object -Last 50 | ForEach-Object { Write-Info " $_" }
	}
	exit 1
	}
	Write-Info "Emulator process started (PIDs: $emulatorProcs)"
	}

[Copilot]

$preferredVersions includes "iOS-26" and the list order contradicts the comment "newest first" (since 26 would be newest but is last). If iOS-26 is a typo, replace it with the intended version; otherwise reorder the list (newest→oldest) so selection matches the comment and expected simulator availability.

Suggested change

	$preferredVersions = @("iOS-18", "iOS-17", "iOS-26")
	$preferredVersions = @("iOS-18", "iOS-17", "iOS-16")

[Copilot]

This section recommends skipping a try-fix attempt after repeating an "unrelated" build error, but earlier in this doc you require STOP on environment blockers. Repeated unrelated build failures are usually an environment/worktree blocker, so the guidance should be aligned (e.g., treat repeatable unrelated errors as a blocker and STOP/ask user, rather than skipping and continuing).

Suggested change

	3. If it fails again with the same unrelated error, skip this attempt and continue with the next model
	3. If it fails again with the same unrelated error, treat this as an environment/worktree blocker: STOP the try-fix workflow for this PR, do **not** continue with the next model, and ask the user to investigate and fix the environment before proceeding (see "Stop on Environment Blockers").

[PureWeen]

/rebase