[KSchlobohm]

Description

Shorten test feedback without reducing the existing OS/Python matrix or selecting fewer tests:

• Run the full suite with two pytest-xdist workers in each existing Ubuntu, Windows, and macOS job, on Python 3.13 and 3.14. Leave the independent Ruff job unchanged.
• Reduce the installed-adapter test HTTP server's shutdown polling interval to 0.01 seconds. Preserve its per-test temporary root, ephemeral port, and shutdown/socket-close/thread-join cleanup.
• Add pytest-xdist>=3.0 only to the test extra. Add hashed execnet 2.1.2 and pytest-xdist 3.8.0 entries and their provenance to the audit snapshot, without changing existing pins or audit behavior.

These are four files, 16 insertions and two deletions. No production code, test exclusions, workflow restructuring, runner settings, or permissions change.

For a directional reference, the slowest test step in the final fork candidate was 833 seconds, versus a mean of 1,675.4 seconds for the slowest test step in each of five successful upstream main push runs sampled on October 20: 39487888924, 39470755550, 39438133070, 39433509775, and 39318274471. This is about 50% lower, not a controlled speedup estimate: commits, test inventories, runner images, and resolved dependencies differ. The metric excludes queue, setup, and retries and is not end-to-end workflow time.

Testing

• Tested locally with uv run specify --help
• Ran existing tests with uv sync && uv run pytest
• Tested with a sample project (if applicable)

Existing hosted evidence: KSchlobohm/spec-kit#92, source 5a4ff5536876845cf894dc1518ec0ac52f916b3b, actual checkout merge a2ac24a3a6d42312dabcfc665a497631e3f4bd9c. Its four-file patch is reproduced here on upstream main f90f7ddf4fa17b2f3bee427f27339cd4d61b0d1a, without changing the patch's scope or behavior.

The final fork run passed uv run pytest -n 2 in all six cells. Each collected 12,787 tests and passed all 671 installed-adapter cases:

OS / Python	Passed	Skipped	Test-step seconds
Ubuntu / 3.13	12,690	97	414
Ubuntu / 3.14	12,690	97	519
Windows / 3.13	12,144	643	813
Windows / 3.14	12,144	643	833
macOS / 3.13	12,686	101	549
macOS / 3.14	12,686	101	637

The two macOS jobs initially failed to acquire runners and executed no steps. One user-authorized failed-jobs-only retry passed both; the original four Ubuntu/Windows passes were retained. There were no executed test failures.

Fork Ruff, markdownlint, shellcheck, the extension version guard, and both CodeQL analyses passed. Dependency audit passed, including the freshness check actually executing on changed dependency inputs and pip-audit reporting no known vulnerabilities.

Checks on this upstream-based worktree: Windows, Python 3.14.8, uv 0.13.2, pytest 9.1.1, pytest-xdist 3.8.0:

Command	Result
uv sync --extra test --python 3.14.8	Passed; isolated editable install resolves to this worktree
uv run specify --help	Passed
.venv\Scripts\python -m pytest tests\test_security_workflow.py tests\specify_cli\integrations\test_installed_adapters.py -k 'test_security_workflow or test_registered_package_force_replacement_restores_sources_and_config_on_rollback or test_round9_optional_metadata' -n 2 -q	82 passed, including existing success and failure/rollback coverage
.venv\Scripts\python -m pytest tests --collect-only -q	12,904 collected
uvx ruff@0.15.0 check src tests	Passed
uvx --from pip-audit==2.10.0 pip-audit --disable-pip --require-hashes -r .github\security-audit-requirements.txt --progress-spinner off	Passed; no known vulnerabilities
git diff --check	Passed

Those checks were run on upstream base 505781d7cabb5180b7098a7d8371c48456f9d4f6. Before publication, upstream gained two integration fixes. The same four-file change was rebased without conflicts onto f90f7ddf4fa17b2f3bee427f27339cd4d61b0d1a. The focused command above was rerun with or test_check_skips_optional_tools_for_external_ide_adapter added to its -k selection: 83 passed. Collection was rerun with the same command: 12,915 collected. git diff --check origin/main HEAD passed.

The final upstream tree has 128 more collected cases than the fork evidence. The full upstream-based suite has not been run locally; its normal PR CI is still needed. Local dependency resolution also differs from the hosted fork environment, so these focused results are not a repeated six-cell validation.

Earlier non-timing fixture probes passed against both original and changed fixtures for HTTP 200/404 responses, distinct roots/ports, and closed sockets/joined threads after normal and simulated-error exits. An earlier same-module Windows/Python 3.13 comparison passed the same 671 cases with two workers: summed worker teardown time fell from 203.53 to 5.61 seconds (baseline, polling treatment). This sum is not wall time; dependencies also drifted, so it does not establish a controlled whole-job gain.

No new permanent exact-YAML or matrix assertions are included. Existing success/failure tests, fixture probes, and hosted results provide observed evidence, not a permanent regression guard for worker count or performance. Sample-project testing is not applicable: the dependency addition is test-only and no shipped CLI or slash-command behavior changes.

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (fill in the disclosure below)

AI disclosure: Prepared with GitHub Copilot using GPT-6.1 Sol (model ID: gpt-6.1-sol), operating autonomously within a user-directed scope. AI assistance covered code/CI edits, dependency snapshot verification, validation, evidence analysis, and drafting this description. Reasoning effort was not explicitly configured in the task.

[Copilot]

🔵 Needs a closer look

Parallel execution still needs passing full-matrix CI results on the final rebased tree.

No open issues 🎉

What changed in this PR

Shortens Spec Kit’s test feedback without reducing the existing OS/Python matrix or test selection.

Changes:

• Runs the full suite with two pytest workers.
• Reduces test HTTP-server shutdown polling to 0.01 seconds.
• Adds test-only parallelization dependencies and audit snapshot entries.

File	Description
tests/​specify_cli/​integrations/​test_installed_adapters.py	Shortens server polling while preserving cleanup.
pyproject.toml	Adds pytest-xdist to the test extra.
.github/​workflows/​test.yml	Enables two workers in existing test jobs.
.github/​security-audit-requirements.txt	Adds hashed dependency pins and provenance.

_{Review effort: Max}

[Copilot]

🔵 Needs a closer look

The passing fork matrix covers a different test inventory, so suite-wide parallel execution still needs validation on the final upstream-based tree.

No open issues 🎉

_{Review effort: Max}