[leonlaiyc]

Problem / Motivation

Goal: the two subagent cwd tests that check ~ expansion run and pass on the Windows matrix instead of sitting in the strict-xfail list.

test/test_subagent_cwd.py::TestValidateCwd::test_tilde_expanded_in_allowed_roots and ::test_tilde_expanded_in_cwd pin HOME at tmp_path and expect validate_cwd to expand ~ under it. On Windows they fail:

assert "cwd is not under an allowed root: ['~/ws']" == ''
assert 'cwd does not exist or is not a directory' == ''

os.path.expanduser on Windows reads USERPROFILE, not HOME. The suite's home floor pins only KIROCREW_HOME, so ~ resolved to the real user profile and both paths fell outside the test's root. validate_cwd does what it should.

Why it matters

As strict xfails, neither test checks tilde expansion on Windows, where a user's ~/... cwd and allowed roots are expanded by exactly this code. Worse, an unpinned ~ points a test at the developer's real profile.

Not a goal

• validate_cwd itself, or how it expands ~.
• The other test files that pin HOME alone. Most are POSIX-only or pass HOME to a child process, and none of them is failing on Windows today.

What changed (motivation → approach → change)

• Symptom: both tests fail on Windows with the path outside the root.
• Root cause: HOME is pinned and USERPROFILE is not, so ~ is not the test's directory on Windows.
• Change: each test also pins USERPROFILE at tmp_path, as pod_helpers.py, agent_home_isolation_helpers.py and member_essential_context_helpers.py already do beside HOME. On POSIX nothing changes. The two strict-xfail lines are gone from test/windows-expected-failures.txt.
• The two lines sit right after the test_slack_gateway_run.py block that test(windows): burn down the gateway-run xfails by patching the rlimit seam #23781 removes, so whichever of the two merges second needs a one-line rebase of the list; I will do it.

Backwards compatibility

Compatible: test-only; no product code changes.

Removes nothing: two strict-xfail lines go, so these tests now run and assert on Windows, and no assertion is dropped.

Tests

• test/test_subagent_cwd.py and test/test_windows_gap_list.py on a Windows host, with the lines removed: 25 passed on Python 3.13.5 and on 3.12.13.
• flake8, black, check_comment_history pass; a trial merge with test(windows): burn down six path-separator xfails #23777 is clean.

Proof: not provable -- test-only change

Manual verification

N/A — test-only change; the Windows run above is the check, and the Windows matrix runs these tests on this PR.

Related Issues

Fixes #23785

Checklist

• At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
• Existing tests pass and new tests added for new functionality
• Self-review completed; code follows project style guidelines
• Documentation updated (if applicable)
• No secrets, credentials, or internal references in the diff

Contribution License Agreement

🤖 Generated with Claude Code

[github-actions]

First Principles Review (Opus 5.5, fork) — ⏭️ skipped

Revision 2da95d92c12f947e74f114ef70f97d8eff63daae ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Does not block merge.

[github-actions]

UX Review (Opus 5.5, fork) — ⏭️ skipped

Revision 2da95d92c12f947e74f114ef70f97d8eff63daae touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

[github-actions]

Design Review (Opus 5.5, fork) — ✅ PASS

Design-level review of 2da95d92c12f947e74f114ef70f97d8eff63daae via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: this is a test-only fix for the cause, and the Windows matrix runs both tests that leave the xfail list.

[DESIGN-REVIEWED] 2da95d9

Description read

[DESCRIPTION-READ] 54f0d7f22a4569d0cb292c8568c1ecc154f8d8e4ff7f433d410884c4bfccd0dc

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

[github-actions]

Opus 5.5 Review (fork) — ✅ no blocking findings

Reviewed 463343236cf993c6c910bbbca58bee72404efee9 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 4633432

[github-actions]

GPT 6.1 Review (fork) — ✅ no blocking findings

Reviewed 463343236cf993c6c910bbbca58bee72404efee9 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 4633432

[leonlaiyc]

PR Readiness on 463343236c is still pending, waiting on Security Scope Review (not started), and that lane stopped on infrastructure, not on the code:

• Fork Security Scope Review run 40114410361: the generate job's Checkout base (trusted) failed three times at 17:04 UTC with fatal: unable to access 'https://github.com/kirodotdev/KiroCrew/': Failed to connect to github.com port 443 ... Couldn't connect to server, and the job was cancelled at 35 minutes. The publish job then posted the [scope-floor:unsettled] failure ("the generate job did not finish").

Design Review passed for this head, and the Opus 5.5 and GPT 6.1 reviews have no blocking findings. A fork cannot re-run workflows: could a maintainer re-run run 40114410361?

[leonlaiyc]

Rebased onto main @ 118702f4d3; new head 2da95d92c1. The diff is unchanged in substance.

• Conflict was only in test/windows-expected-failures.txt: main has since removed the neighbouring test_slack_gateway_run.py lines, and this PR deletes the two test_subagent_cwd.py lines. Neither set remains, so the PR's diff is still the same two deletions plus the two USERPROFILE pins in test/test_subagent_cwd.py.
• On this Windows host: test/test_subagent_cwd.py 30 passed (the list applies, and no XPASS); test/test_ci_surface_tests.py 74 passed.