Skip to content

Development

m4bwav edited this page Nov 21, 2026 · 1 revision

Layout

Path What it is
.claude-plugin/plugin.json plugin manifest
skills/testwright/SKILL.md the skill
scripts/tw.py the CLI, standard library only
references/ one file per stack, plus quality, legacy and games
tests/test_tw.py, tests/fixtures/ unit tests for the parsers, lint and diff reading
evals/evals.json, evals/fixtures/ five eval cases with no-skill baselines
evergreen.json, RESEARCH.md, LEARNINGS.md, TESTS.md, CHANGELOG.md evergreen upkeep files

Tests

There is no CI; tests run locally:

python -m pytest -q tests/test_tw.py

At v0.1.0 this reports 10 passed.

Point pytest at tests/test_tw.py, not tests/: tests/fixtures/bad_test.py is a lint fixture with deliberately broken tests, and pytest collects it from tests/ (one failure, a NameError).

Evals

evals/evals.json holds five cases (planning, a red suite, a flaky suite, weak tests, legacy code), each run with and without the skill. Results are in TESTS.md.

Releasing

Bump the version in .claude-plugin/plugin.json and in the skill's metadata.version, add a CHANGELOG entry, tag vX.Y.Z and cut a GitHub release. Then update this wiki, whose footer names the version.

Clone this wiki locally