# v0.3.8-alpha.3 Release Notes

## Version

`v0.3.8-alpha.3`

## Dist Tag

`alpha`

During the alpha pilot, `latest` may be synchronized to the current alpha build to avoid stale default installs. This is not a stable support claim.

## Summary

This alpha is a P0/P1 workflow-guard hardening release for the v0.3.8 line.

It focuses on making Windows/OpenCode runs safer and easier to judge before release:

- Windows doctor/runtime readiness detection is more robust;
- generated backend stack mismatch can block workflow finish;
- fake Java/Spring shortcuts are called out more directly;
- generated Java role read coverage and backend-shape review evidence are surfaced;
- init/bootstrap docs and CLI output now describe the actual setup split;
- Windows bearshell/read/search guidance avoids unsafe vendor-recursive searches.

This release still does not certify full OpenCode implementation completion or generated product quality.

## Added

- Windows-aware OpenCode command detection for `ph doctor`.
- Stack-alignment profile checks for Java/Spring/Gradle backend output.
- `ph workflow finish implement` blocking when stack mismatch evidence is present.
- Java role read coverage checks for generated backend files.
- Backend-shape review report status.
- Windows-safe workflow read guidance.
- Windows bearshell quoting guidance for PowerShell-oriented commands.
- Bounded, vendor-safe Windows search guidance for README/project-scoped searches.

## Changed

- Java/Spring backend guidance now forbids fake Gradle shims and Java `HttpServer`/CommonJS-style workarounds for generated backend projects.
- `ph init` is documented and reported as the minimal harness/OpenCode integration step.
- `ph bootstrap backend` is documented and reported as the backend-ready path that prepares `AGENTS.md`, backend profile, policy overlay, accepted plan, report templates, harness config, and OpenCode config.
- Windows bearshell search guidance now prefers bounded project files such as `README.md` instead of recursive vendor-directory scans.

## External Smoke

Windows P0/P1 smoke result: partially successful.

Observed:

- `ph doctor` passed.
- Runtime readiness passed.
- `ph init`/`ph bootstrap backend` output matched the generated artifacts.
- `profileSummaryInjected` was confirmed.
- Backend-shape report generation was observed.
- Fake Gradle shim, Java `HttpServer`, and CommonJS workaround paths were not observed.

Not completed:

- Full OpenCode finish did not complete cleanly inside the smoke window.
- The run timed out or continued with pending work, so full implementation completion remains unverified.
- Generated app product quality is not certified by this release.

Windows vendor-safe bearshell search result: successful.

Observed:

- README-only `Select-String -Path README.md -Pattern TODO` guidance passed.
- The corresponding command output and command execution passed.
- Unsafe recursive/vendor search guidance was not observed.

## Verification

Repository verification from the release prep worktree:

```bash
npm test
npm run typecheck
npm run build
npm pack --dry-run
npm run check:docs
```

Observed result on 2026-06-25:

- `npm test`: passed. Scope diagnostics PASS, docs taxonomy diagnostics PASS, 55 files / 337 tests.
- `npm run typecheck`: passed.
- `npm run build`: passed.
- `npm pack --dry-run`: passed for `persona-harness@0.3.8-alpha.3`.
  - Filename: `persona-harness-0.3.8-alpha.3.tgz`
  - Total files: 328
- `npm run check:docs`: passed.

## Known Gaps

- Windows P0/P1 smoke is partially successful, not a full success.
- Full OpenCode implementation finish remains incomplete because the smoke timed out or continued with pending work.
- This release does not certify generated app product quality.
- Workflow guards improve observable failure and guidance surfaces, but they are not a replacement for external full-run validation.

## Recommended Before Publish

- Publish only as a prerelease alpha.
- Keep release messaging focused on workflow guard, Windows guidance, and evidence/reporting improvements.
- Do not describe the Windows smoke as full implementation success.