testing: C#/.NET idioms are the fallback defaults and only worked examples in ecosystem-agnostic skills

[kyle-sexton]

Plugin: testing · Categories: hardcoded opinion (1), externalization gap (2)
Source: work-readiness sweep (read-only audit vs docs/PLUGIN-PHILOSOPHY.md + docs/MIGRATION-PLAYBOOK.md)

Findings

• plugins/testing/skills/write/SKILL.md:41 — fallback naming defaults {Method}_Should{Behavior}_When{Condition} etc. (PascalCase C# idiom presented as THE universal default, unqualified); also :59-60
• plugins/testing/skills/write/context/organize.md:47-56 — "Test class: {ClassUnderTest}Tests" under a bare "## Naming" heading, no ecosystem qualifier
• plugins/testing/skills/write/context/write.md:95-113 — the only worked code sample is a C#/xUnit [Fact] block; also :47-53,119-121
• plugins/testing/skills/diagnose/context/loop.md:59-65 — the sole regression example is dotnet test
• Also: plugins/testing/skills/plan/SKILL.md:39-42, run-e2e/context/non-ui.md:22-29,73-79, diagnose/SKILL.md:49-50, diagnose/context/investigate.md:15,17,21

Why it breaks agnosticism

The skills hedge correctly elsewhere ("comes from the consuming project's testing conventions") — real mitigation. But the FALLBACK DEFAULTS a consumer lands on when nothing is documented, the only worked code sample, and the regression command are uniformly C#/.NET with no neutral counterpart. A Python/Go/JS/Rust consumer gets house-stack-only illustration as canonical. The seam exists; the default it falls back to is the issue.

Fix direction

State fallback defaults ecosystem-relative and route through the convention-resolution ladder ("use the project's documented pattern; when undocumented, mirror the ecosystem's idiom") instead of a PascalCase C# template as universal. Keep .NET examples but pair the sole code sample/regression command with one non-.NET idiom or label the block "illustrative (.NET/xUnit)". Demote house-stack forms from "default" to "one labeled example" — not a hardcode swap.

[kyle-sexton]

This was generated by AI during triage.

Triage → T2 planning complete · confidence: high (advisor-reviewed). All 8 cited files accounted for (6 edited, 5 verified already-relative; one file in both buckets). Framing/labeling only — no behavioral guidance moves. Labeling status: ready.

Fix plan — demote C#/.NET fallback defaults to ecosystem-relative in the agnostic skills

Behavioral contract

Framing and labeling only. TDD cadence, Four Pillars, verify-through-the-interface, the reproduce→isolate→fix→retest→regression loop, and all routing/handoff stay byte-identical. One rule everywhere a house-stack form currently reads as THE default:

Convention-resolution ladder — use the project's documented pattern; when undocumented, mirror the consuming ecosystem's idiom (never impose one language's convention on another).

Concretely: (a) reword every "default" into the ladder, (b) demote each PascalCase C# form from "default" to one labeled illustrative (.NET/xUnit) example, (c) for the sole worked code sample and the sole regression command, add an "illustrative (.NET)" label. We do NOT mint a parallel Python/JS default set — that re-commits the same sin in a new language.

Per-site edits (required — 6 sites)

1. skills/write/SKILL.md — "Naming" bullet. Reword the unqualified defaults into the ladder + label the forms "illustrative (.NET/xUnit) — adapt casing/separators to the target ecosystem."

2. skills/write/context/write.md — Step 3 naming block. Replace the lead line so the three name-form bullets are prefaced "Use the project's documented naming pattern; when undocumented, mirror the ecosystem's idiom. The forms below are illustrative (.NET/xUnit)…"; bullets unchanged.

3. skills/write/context/write.md — the ONLY worked code sample. Label the [Fact] block's lead-in "…Verify through the same interface callers use (illustrative — .NET/xUnit; the principle is ecosystem-agnostic):". Code block unchanged. (Label, not pair — pairing would author a second worked example, new surface to get wrong.)

4. skills/write/context/organize.md — bare "## Naming" form. "Test class: {ClassUnderTest}Tests" → "the project's documented test-class naming; when undocumented, mirror the ecosystem's idiom ({ClassUnderTest}Tests illustrates the .NET convention)".

5. skills/diagnose/context/loop.md — the sole regression command. Reword the prose line to route through the project's runner ("/toolchain:check is SSOT for the exact per-ecosystem command") and label the dotnet test block illustrative (.NET). Command block unchanged — no new command string minted.

6. skills/plan/SKILL.md — Step 2 output-template placeholders. Preface the template with "Test-name forms follow the project's documented pattern; when undocumented, mirror the ecosystem's idiom — the PascalCase placeholders below are illustrative (.NET/xUnit)." Template block unchanged.

Cited files verified ALREADY ecosystem-relative (no change)

• skills/write/SKILL.md Gotchas — ".NET: xUnit v3 rejects --nologo…" already labeled framework trap.
• skills/diagnose/SKILL.md — "Framework traps — .NET examples: …" already labeled.
• skills/diagnose/context/investigate.md — already pairs idioms (--filter xUnit; -k pytest; --testNamePattern vitest). Model site to emulate.
• skills/run-e2e/context/non-ui.md — "vitest for Node, xUnit for .NET, pytest for Python" already paired.
• skills/write/context/organize.md table/Test-project line — already hedged "(e.g. {Project}.Tests… in .NET)". (Only the bare "## Naming" line, edit ci: onboard to the CI platform (ci-workflows + standards) #4, needs work.)
• skills/plan/SKILL.md classification table — "(e.g. WebApplicationFactory for .NET)" already hedged.

Acceptance criteria

• No unqualified PascalCase form remains presented as THE universal default; each routed through the ladder + labeled illustrative (sites 1, 2, 4, 6).
• The sole worked code sample carries an "illustrative — .NET/xUnit" label (site 3).
• The sole regression command labeled illustrative (.NET) + routed through /toolchain:check (site 5).
• Diff shows only lead-in/framing prose; code and template blocks byte-identical.
• All 8 cited files accounted for (6 edited + already-relative sites above).

Notes

• .NET refs in write/diagnose evals.json are .NET-scenario fixtures (a .NET prompt exercising the skill), not fallback defaults — out of scope here (that leak is testing: evals grade against the private source monorepo's layout and house test stack (+ ghost testing.md ref) #428).
• Deliberately avoided stacking a parallel Python/JS default set — that reproduces the blanket-hardcode-swap the issue forbids.