Fix Windows test suite compatibility

[JJM1986]

Windows test execution exposed two portability issues that prevented the suite from running reliably. This change keeps pytest's generated environment metadata below Windows' 32,767-character limit and makes the excluded-artifact size assertion account for platform newline encoding.

Validation:

• 158 focused tests passed, 10 skipped
• Ruff lint and format checks passed
• Full non-integration validation progressed beyond the original failures; unrelated pre-existing Windows-specific failures remain in other suites

No source behavior was changed; the changes are limited to test portability and deterministic parametrization.

[rng1995]

[SkillSpector Review]

Hi @JJM1986, thank you for your contribution to SkillSpector — we really appreciate the time you put into this! A few items need attention before it can be merged; details below.

This PR makes two Windows test-portability fixes: explicit pytest ids for the long-string parametrization in tests/nodes/analyzers/test_json_container_ownership.py, and a newline-aware size assertion in tests/nodes/test_build_context.py. Its second commit (101c82ab, "feat(cli): add offline Runway project exporter") also adds an unrelated production feature: a skillspector runway-project CLI command, a new skillspector.runway_projects module, a README section and tests. I checked the two test fixes by reading the code, and both are correct. All 19 ids match the 19 parameter sets. _MAX_JSON_QUOTE_CONTAINER_CHARS is 65,536, so the auto-generated test ID would exceed Windows' 32,767-character limit when pytest sets PYTEST_CURRENT_TEST. write_text() writes CRLF on Windows, so the os.linesep expectation is right. Both changes behave the same as before on POSIX. The Runway commit is the blocker.

Findings

1. [Blocking] An unrelated feature is bundled into this test-portability PR (src/skillspector/cli.py:3092, src/skillspector/runway_projects.py:1, tests/unit/test_runway_projects.py:1, README.md:147).

   • The new command exports prompt sheets for Runway video generation from hard-coded templates ("The Legends of Automation", "Cerberus Wolf: Drone Runaway"). The templates cite an "attached video brief" that is not in the repository.
   • It is unrelated to both Windows compatibility and skill scanning, and there is no linked issue.
   • It adds a public CLI command and shipped package code to a security tool. This contradicts the PR description ("No source behavior was changed; the changes are limited to test portability").
   • Existing commands are unaffected, because the change only adds a command and an import. That is not a reason to include it here.
   • The new files lack the SPDX header that CONTRIBUTING.md requires for new source files.
   • The README section is inserted just before the existing #### Size limits subsection, which now reads as part of the Runway section.

   Please drop commit 101c82ab, for example by resetting the branch to 6a8f6941 and force-pushing, so the PR contains only the Windows test fixes. If you want to propose this feature, please open an issue first.

2. [Non-blocking] tests/nodes/test_build_context.py:1378: optional. The new expectation is correct, but the test file's contents still depend on the platform. Writing it byte-exact (payload.write_bytes(b"ordinary metadata\n")) and keeping len(b"ordinary metadata\n") would make the test identical on every platform.

3. [Non-blocking] tests/nodes/analyzers/test_json_container_ownership.py:89: optional. A separate ids=[...] list can silently mislabel cases if entries are reordered later. pytest.param(value, id="...") keeps each ID next to its case.

Tests/CI: All checks pass at 101c82ab (DCO, lint, test-unit, docker-smoke, OpenCode). CI runs only on Ubuntu with Python 3.12, so the Windows behavior itself is not exercised. Related: #636 (same author) makes separate Windows fixes under scripts/ and does not touch the same files. Once the Runway commit is removed, the remaining two test changes look ready to approve.

---

Decision: Changes Requested (reviewed head 101c82ab02ccffbfce7119e31e7ba1c1a60e3a31)

[rng1995]

[Blocking] This runway-project command generates Runway video-prompt sheets from built-in templates. It is an unrelated product feature, as are the skillspector.runway_projects import and the RunwayFormat enum it needs, and none of it belongs in a Windows test-suite fix. It adds a public CLI command to a security scanner without an issue or design discussion. It also contradicts the PR description ("No source behavior was changed"). Please drop commit 101c82ab from this PR.

[rng1995]

[Blocking] This new module ships in the package but has nothing to do with scanning skills. It contains hard-coded "Cerberus Wolf" and "Legends of Automation" video prompts that cite an "attached video brief" not present in the repository. It also lacks the SPDX header that CONTRIBUTING.md requires for new source files. Please remove it together with tests/unit/test_runway_projects.py.

[rng1995]

[Blocking] Please remove this section along with the rest of the Runway feature. Where it sits now, it also splits the Basic Usage docs. The existing #### Size limits subsection, which documents the caps on remote and archive inputs, now appears as a child of "Runway video projects".

[rng1995]

[SkillSpector Review]

@JJM1986 thank you for running the suite on Windows and fixing what you found.

Value and readiness

This is a status re-review of 101c82a against current main (530d3c5). The branch has not changed since the 28 September review, and it merges cleanly into current main. main has not picked up either Windows fix, so both are still useful:

• the 19 ids still match the 19 parametrized cases;
• _MAX_JSON_QUOTE_CONTAINER_CHARS is still 65,536;
• the merged test_build_context.py still writes its payload with write_text.

The only blocker is still the unrelated Runway commit. Once that commit is dropped, the two test fixes are ready to approve.

Material findings

Prior findings:

1. Blocker: commit 101c82ab adds an unrelated runway-project command, the runway_projects.py module, a README section and tests (src/skillspector/cli.py:3092, src/skillspector/runway_projects.py:1, README.md:147; open threads). Still open. Please reset the branch to 6a8f6941 and force-push.
2. Optional: write the test payload byte-exact with write_bytes in tests/nodes/test_build_context.py. Still open.
3. Optional: use pytest.param(..., id=...) instead of a separate ids list. Still open.

I found nothing new.

PIC tradeoffs

None identified.

Verification and gaps

• git merge-tree against 530d3c5 is clean. In the merged tree the parametrize decorator has 19 cases and 19 ids, and import os is present for the os.linesep assertion.
• CI passed at this head on 25 September (DCO, lint, test-unit, docker-smoke, OpenCode). That run used an older main, so CI should re-run after the branch is updated. CI runs only on Ubuntu, so it does not exercise Windows.
• The author has not responded on the PR in the 15 days since the review.