Bd companion plugin
Skill Lonsdale201/wp-agent-skills/better-data/bd-companion-plugin
A community-maintained collection of agent skills for WordPress plugin and theme development.
npx -y skills add Lonsdale201/wp-agent-skills --skill bd-companion-pluginAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 21 stars21 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Work on better-data-plugin-test — the companion plugin that exercises the better-data library against a live WordPress install. Plugin is intentionally NOT part of the library public API, so feel free to break its internals to demonstrate a point. Three test tiers — Smoke (regression; never tolerate FAIL), Stress (deep integration with OK/FAIL/NOTE findings — NOTE is for surfaced quirks worth documenting without blocking), and Admin pages (eyeball-level proof of behaviour, e.g. ShopSettingsPage rendering print_r($dto) to visually confirm Secret redaction). The Widget Shop fixture (bd_widget CPT + bd_order CPT + ShopSettingsDto) is the canonical realistic consumer; extend it rather than inventing a new fixture. CLI is the main driving surface — wp better-data {test, stress, seed, purge, inventory}. Use when changes go under wp-content/plugins/better-data-plugin-test/. Triggers on Smoke / Stress / Runner / Cli files in that path, "wp better-data" CLI invocations, "smoke / stress scenario" mentions.
SKILL.md
14.4 KB, as published. Nobody here has run it
better-data: Companion plugin testbed
For library maintainers working in the integration testbed at wp-content/plugins/better-data-plugin-test/. The plugin's job is to verify better-data behavior against real WordPress — things that pure unit tests can't reach (cache primings, metadata_exists semantics, wp_slash round-trips, REST schema actually appearing in WP_REST_Server, locale switching, encryption key constants).
Misconception this skill corrects
"I'll add the new feature's integration test directly inside
better-data/tests/Unit/so everything's in one place."
Wrong. tests/Unit/ is the WP-free zone — every test there must run with composer test against a clean PHP environment, no WP bootstrap, no MySQL. That's intentional: contributors can run the unit suite in seconds. WP-aware behavior goes in the companion plugin, where there's a real WordPress to talk to.
The split:
| Concern | Lives in |
|---|---|
| Pure type coercion, attribute reflection, builder logic | better-data/tests/Unit/ |
wp_slash round-trip, metadata_exists semantics, REST registration | better-data-plugin-test/src/Smoke/ |
| Multi-step scenarios (encrypt → store → fetch → decrypt → reveal), edge-case discovery | better-data-plugin-test/src/Stress/ |
Visual confirmation (admin page renders print_r($dto)) | better-data-plugin-test/src/Admin/ |
Other AI-prone misconceptions:
- "I'll let the plugin depend on WooCommerce since it makes the Order fixture realistic." Wrong — the plugin must run on a clean WP install with NOTHING but better-data. WC, ACF, custom-fields plugins are out. The Widget Shop fixture (
bd_widgetCPT +bd_orderCPT +ShopSettingsDto) IS the realistic consumer because it's self-contained. - "If a stress scenario is flaky, I'll mark it as
NOTEinstead of fixing it." Wrong —NOTEis for documented quirks the library legitimately surfaces (e.g. "WP serializes integer-keyed arrays as objects in some contexts"). A flaky test is aFAILwaiting to happen; fix it. - "I'll move the new behavior into the library after the plugin verifies it." Wrong direction — integration-only behavior STAYS in the plugin. The library stays shape-agnostic.
When to use this skill
Trigger when ANY of the following is true:
- The diff modifies any file under
wp-content/plugins/better-data-plugin-test/. - Adding a new smoke or stress scenario.
- Adding a new CLI subcommand under
bin/wp better-data <subcommand>. - Adding a new admin page, fixture DTO, or seed/purge routine.
- Reviewing a PR that adds WordPress / WC / ACF as a Composer dep on the plugin.
Workflow
1. Three test tiers — pick the right one
Smoke — regression coverage.
src/Smoke/
├── Runner.php ← scenario list + dispatcher
└── Assertion.php ← per-scenario assertions
Smoke scenarios run on every library change. They're short, focused, fast. A FAIL here means a regression — never ship over it. Add a smoke scenario for any new public behavior.
Stress — deep integration.
src/Stress/
├── Runner.php ← scenario list + dispatcher
└── Finding.php ← OK / FAIL / NOTE findings
Stress scenarios are longer-running, multi-step, edge-case-hunting. They produce three outcomes:
| Verdict | Meaning |
|---|---|
OK | Library behaves as expected for this scenario |
FAIL | Bug — blocks the change |
NOTE | Discovery — library surfaces a quirk worth documenting (e.g. "PHP arrays with integer keys round-trip as JSON objects in this context") |
NOTE is the unique value-add: it lets a stress run discover and SURFACE library limits without failing.
Admin pages — visual confirmation.
src/Admin/
├── AdminPage.php ← base
├── ShopSettingsPage.php ← renders print_r($dto) post-hydration
├── …
Admin pages are eyeball-level proof. Use for:
- Visually confirming
Secretrenders as'***'. - Confirming locale-switched email subject lines render in the right language.
- Manually playing with
Presenterchains.
Not test scaffolding — they're for human verification.
2. The Widget Shop fixture is the canonical consumer
The plugin defines:
bd_widgetCPT — products in the fictional shop.bd_orderCPT — orders.ShopSettingsDto— store-level settings (currency, tax, encrypted Stripe key).WidgetDto,LineItemDto,OrderDto,CustomerDto,AddressDto— see src/Dto/.
When you add a new feature:
- Can it fit into the Widget Shop? (Yes for most things — sinks, sources, validation rules, presenter methods.) Extend the existing fixtures.
- Does it genuinely need a new fixture? (E.g. testing a new SOURCE that doesn't fit the shop concept.) Add a fixture, but keep the dependency-tree minimal.
Adding a new CPT just for one test is over-fitting. Adding a new DTO that consumes existing CPTs is fine.
3. CLI subcommands
The plugin's main driving surface is CLI:
wp better-data smoke # run all smoke scenarios
wp better-data stress # run all stress scenarios
wp better-data seed # populate the Widget Shop with sample data
wp better-data purge # clear all bd_* posts and meta
wp better-data inventory # list every DTO + scenario the plugin exposes
Existing files: src/Cli.php, src/StressCli.php, src/SeedCli.php, src/PurgeCli.php, src/InventoryCli.php.
Add a new subcommand when you have a coherent chunk of work that doesn't fit the existing five (e.g. bench for benchmark scenarios). Don't add one for a single scenario — that goes inside an existing tier.
4. Adding a stress scenario
// src/Stress/EncryptionRotationScenario.php
namespace BetterDataPluginTest\Stress;
final class EncryptionRotationScenario
{
public function run(): array // list<Finding>
{
$findings = [];
// 1. Set up: define BETTER_DATA_ENCRYPTION_KEY, encrypt + store a Secret.
// 2. Define BETTER_DATA_ENCRYPTION_KEY_PREVIOUS = old key.
// 3. Read back: should still decrypt under previous key.
// 4. Re-write: should encrypt under new key.
// 5. Read back: under new key.
// 6. Verify the wp_options row's stored value uses the new key envelope.
$findings[] = Finding::ok('round-trip under rotation works');
$findings[] = Finding::note('previous-key rows are not auto-rewritten on read; lazy migration on write only');
return $findings;
}
}
Wire it into the runner:
// src/Stress/Runner.php
public function scenarios(): array
{
return [
new RoundTripScenario(),
// ...
new EncryptionRotationScenario(),
];
}
5. Adding a smoke scenario
Smoke is more compact — assertion-style, no Finding objects:
// src/Smoke/Runner.php — inside the scenario list:
$assertions->ok(
'PostDto round-trips through PostSink + PostSource',
fn () => $this->postRoundTrip(),
);
The assertion closure throws on failure or returns nothing on pass. Single-line scenario titles, single-step assertions.
6. Don't depend on WC / ACF / external plugins
The companion plugin's composer.json lists ONLY:
php: ^8.3lonsdale201/better-data(path repository pointing back to../../libraries/better-data)
No WooCommerce, no ACF, no Composer packages outside the library. Reason: a contributor cloning the library + plugin should be able to run the smoke + stress suite on a clean WP install, no setup choreography. If the plugin grew a dep on WC, half the contributors couldn't reproduce.
7. Run order
# In the library:
vendor/bin/phpunit # unit suite — must be green first
vendor/bin/phpstan analyse # static analysis
vendor/bin/php-cs-fixer fix # style
# In the plugin (against a real WP):
wp better-data purge # clean slate
wp better-data seed # populate the Widget Shop
wp better-data smoke # regression
wp better-data stress # deep integration
# Iterate on FAIL findings until clean.
A successful sequence is the ship-readiness check. Library unit + static + style green, plugin smoke 100% pass, stress 0 FAIL.
Critical rules
- Plugin is internal — break its internals as needed. Not part of the library's public API; redesign freely to demonstrate a point.
- Three test tiers, pick the right one. Smoke = regression (never FAIL); Stress = integration (FAIL blocks, NOTE documents); Admin pages = visual confirmation.
NOTEis for documented quirks, not flaky tests. A test that flakes is a FAIL waiting to happen — fix it.- Widget Shop is the canonical fixture. Extend it; add new fixtures only when the new use case doesn't fit.
- No deps beyond
better-dataitself. Plugin must run on a clean WP install. No WC, no ACF, no third-party Composer packages. - CLI is the driving surface.
wp better-data {test, stress, seed, purge, inventory}covers the workflow; add a subcommand only for genuinely new categories. - Integration-only behavior STAYS in the plugin. Don't promote a stress-only feature into the library.
- Smoke green + 0 FAIL stress = ship-ready. That's the gate.
Common mistakes
// WRONG — depending on WooCommerce in the plugin
// composer.json: "require": { "woocommerce/woocommerce-stubs": "^8.0" }
// Now contributors need WC installed to run the suite.
// RIGHT — keep the dep tree to better-data + WP
// WRONG — scenario marked NOTE because it's flaky
$findings[] = Finding::note('Sometimes the cache primes too late and we read a stale value');
// WRONG: a flake is a FAIL waiting to happen.
// RIGHT — fix it (add explicit cache priming, or design the scenario to be deterministic)
$findings[] = Finding::ok('cache primes correctly when explicitly warmed');
// WRONG — new CPT for one test
\register_post_type('bd_email_log', [...]); // just to test EmailLogSink
// Bloats the fixture surface for everyone.
// RIGHT — extend the Widget Shop
// EmailLogDto becomes part of the bd_order CPT meta, or the bd_widget reviews flow.
// WRONG — promoting a plugin-only behavior into the library
// "OrderTotals helper grew useful — moving it into better-data/src/Helpers/"
// Library stays shape-agnostic; helpers are a consumer concern.
// RIGHT — keep it in the plugin
// WRONG — skipping wp better-data purge before stress
// Stale seed data from a previous run skews scenarios.
// RIGHT
wp better-data purge
wp better-data seed
wp better-data stress
// WRONG — putting integration-test code in the library
// better-data/tests/Integration/EncryptionRoundTripTest.php (requires WP bootstrap)
// Breaks the WP-free unit suite contract.
// RIGHT
// better-data-plugin-test/src/Stress/EncryptionRoundTripScenario.php
Cross-references
- Run
bd-data-objectwhen adding a new fixture DTO undersrc/Dto/. - Run
bd-source-adapter+bd-sinkwhen the scenario covers a new WP store integration — the plugin is where you verify cache primings work. - Run
bd-securityfor stress scenarios involvingSecret/#[Encrypted]— the plugin is where leak probes against live WP run.
What this skill does NOT cover
- Plugin distribution / packaging. The companion is repo-internal; it's not published.
- WordPress bootstrapping for the unit suite. Library unit tests must run WP-free.
- Performance benchmarking with realistic scale (10k posts, 1M meta rows). Stress is correctness-focused; benchmarks would be a separate
benchsubcommand. - Plugin-level i18n. Plugin uses simple English strings; libraries' i18n behavior is in the library's tests.
- Admin UI design / UX. Admin pages exist for visual confirmation, not to demonstrate good UI.
- Multi-site behavior. WP multisite is an edge case for some sinks/sources; covered case-by-case in stress, not as a tier.
References
- Plugin entry: wp-content/plugins/better-data-plugin-test/better-data-plugin-test.php — bootstrap, CPT registration, CLI hookup.
- Widget Shop fixtures: wp-content/plugins/better-data-plugin-test/src/Dto/ —
WidgetDto,OrderDto,LineItemDto,CustomerDto,AddressDto,ShopSettingsDto,EncryptedCredentialsDto,OptionalBeforeRequiredTrapFixture. - Smoke runner: wp-content/plugins/better-data-plugin-test/src/Smoke/Runner.php and src/Smoke/Assertion.php.
- Stress runner: wp-content/plugins/better-data-plugin-test/src/Stress/Runner.php and src/Stress/Finding.php —
Finding::ok,Finding::fail,Finding::notefactories. - CLI: wp-content/plugins/better-data-plugin-test/src/Cli.php, src/StressCli.php, src/SeedCli.php, src/PurgeCli.php, src/InventoryCli.php.
- Admin pages: wp-content/plugins/better-data-plugin-test/src/Admin/ —
AdminPagebase + concrete pages. - Official documentation: https://github.com/lonsdale201/better-data
- Verified source paths:
wp-content/plugins/better-data-plugin-test/src/Plugin.php