[Marius Storhaug (MariusStorhaug)]

Testing guidance prioritizes observable supported behavior, real implementations, and safe actual APIs instead of assertions that merely detect implementation edits. A fast credential-free local loop remains available; protected live checks carry explicit prerequisites, authorization, isolation, cleanup, and honest coverage evidence.

The shared plugin adds a thin testing-skill pointer. This is a standards/plugin-source delivery, not a test-suite migration or consumer rollout: existing installations are unchanged, and the focused skill requires updating the shared msx plugin after this source is merged and available.

New: A focused skill routes test work to the canonical policy

msx-coding-testing routes test design, authoring, and review to the single Testing standard. The skill copies no policy. Consumers can continue using their existing general coding route; to gain the focused trigger, update their shared msx plugin after the source is available. This PR does not install it for them.

Technical details

The actual plugin and marketplace entry advance together from 0.1.1 to 0.1.2. The independently managed marketplace metadata stays 0.1.0. Existing plugin tests pass (4 tests, 49 skill directories), both changed plugin Markdown files lint cleanly, and the new pointer resolves when explicitly passed to the existing link tool (1 target in 1 file).

Changed surface	Standards checked	Framework docs checked	Result
.github/plugin/msx/skills/msx-coding-testing/SKILL.md	Documentation Model, Natural Language, Markdown	Agent Plugin Marketplaces design; sibling thin-skill pattern	Aligned
.github/plugin/msx/plugin.json and .github/plugin/marketplace.json	Manifest identity and synchronized versions	Agent Plugin Marketplaces design; plugin README	Aligned
.github/plugin/README.md	Documentation Model, Natural Language, Markdown	Agent Plugin Marketplaces design	Aligned

Changed: Testing policy distinguishes functional evidence from edit detection

Contributors and agents choose tests around supported functions, commands, and APIs without an arbitrary mocked-unit-test ratio or unnecessary remote calls for pure logic. Mocks and fakes are justified exceptions with explicit limits; a fake or an unexecuted live check does not prove that an external API works. A behavior-preserving refactor must not fail tests solely because implementation text changed.

The engineering principle links to the owning policy instead of requiring a cloud-free full suite. Independent expected results, meaningful examined-unit counts, regression intent, linting, formatting, static analysis, link checks, builds, and temporary-fixture tooling checks remain valid. Production writes and unsafe, destructive, or paid calls gain no authorization from this standard. No runtime, CI, or existing-suite migration is included.

Technical details

Scoped comparison: MSXOrg/docs base aa9b5cecc69f4cbe6aa04a5dd70ff9cdadeeac3d to immutable candidate 607cca0010fa08b9fee43c6f5b07c17c5a051253, in two focused ordinary commits. The complete diff contains seven files; no tests, scripts, workflows, runtime dependencies, or settings-delivery changes are imported. The worktree is clean.

Changed surface	Standards checked	Framework docs checked	Result
src/docs/Coding-Standards/Testing.md	Existing standard shape, Documentation Model, Natural Language, Markdown	None (no framework-specific docs)	Fixed in this PR
src/docs/Ways-of-Working/Principles/Engineering-Practices.md	Documentation Model, Natural Language, Markdown	Owning Testing policy	Fixed in this PR
src/docs/Coding-Standards/index.md	Generated index and recursive-navigation conventions	Repository README and contribution instructions	Aligned

Validation on the committed tree:

• Existing Pester runner: 80 passed in 7 suites; zero failed, skipped, inconclusive, or unexecuted.
• Markdown lint: all 5 changed Markdown files, zero issues.
• Internal links: 127 documentation files scanned, all resolve.
• Cross-repository links: 7 targets in 131 files, all resolve; the new skill pointer was checked separately as recorded above.
• Generated index verification and diff whitespace checks pass.
• Clean site build: repository-pinned Zensical 0.0.59, restored in an isolated environment, no issues. The global 0.0.50 mismatch was detected and is not the pinned build evidence; no repository dependency or shared tool was changed.
• Exact-head Docs CI: Test, Lint, Build, and Links passed; Publish skipped for the PR. Exact-head cross-repository CI passed.
• Exact-head Copilot review: approved at 2026-11-02T18:38:24Z, covering all seven changed files. The explicit review loop returned clean with zero new comments; complete review-thread pagination returned no threads.

The scoped Task records baseline/scenario reasoning, both commits, local checks, CI/review evidence, and completed standards/convergence review. Every implementation-plan item is complete; the Task remains open for its normal reviewed-merge completion path. Targeted open issue/PR searches for testing, mocks, and functional tests found no additional fully convergent issue; the settings and release deliveries remain separate. No policy-wording or structural-content tests were added, and no existing suites were changed or deleted.

Native Agent merge was verified before enabling and read back afterward: enabled=true, address_reviews=true, fix_ci=true, resolve_conflicts=true, merge_pr=false. This is maintenance-only authorization. The PR remains draft by request with GitHub auto-merge unset; advancing readiness and any later merge require the appropriate user/owner decision. No merge, enqueue, service publication, permission/rule/label change, or consumer rollout was performed or authorized. No protected live-API experiment campaign was run or claimed. The inspected Docs route only publishes Pages on non-PR events and does not invoke Release Management; no release-decision block or labels are applied.

Related references

• Resolves Define behavior-focused testing and a thin testing skill #234
• References Solution settings use canonical filenames #233 (separate settings delivery; unchanged)

[Copilot]

🟢 Approved

No active review issues are recorded for the testing-guidance and plugin-skill changes.

No open issues 🎉

What changed in this PR

This PR makes the Testing standard the single owner of behavior-focused testing guidance and adds a plugin skill that points agents to it. It does not migrate existing suites or install the skill into consumers.

Changes:

• Prioritizes supported behavior, real implementations, independent expected results, and justified test doubles.
• Preserves a credential-free local loop while defining authorization, isolation, cleanup, and evidence requirements for protected live-API checks.
• Adds the testing skill pointer and synchronizes plugin versions to 0.1.2.

File	Description
src/​docs/​Ways-of-Working/​Principles/​Engineering-Practices.md	Delegates local-testing requirements to the Testing standard.
src/​docs/​Coding-Standards/​Testing.md	Defines behavior-focused testing policy and protected live-check requirements.
src/​docs/​Coding-Standards/​index.md	Updates the Testing standard’s summary.
.github/​plugin/​README.md	Includes testing in the coding-skill scope.
.github/​plugin/​msx/​skills/​msx-coding-testing/​SKILL.md	Adds a pointer to the canonical Testing standard.
.github/​plugin/​msx/​plugin.json	Bumps the plugin version to 0.1.2.
.github/​plugin/​marketplace.json	Aligns the marketplace version with the plugin.

_{Review effort: Balanced}