[Blind-Striker]

📝 Description

What does this PR do?
It applies the slice-home rule to OpenCode.Sdk.Tools.Tests, the second half of #402. Helpers and embedded fixtures move into the slice whose code uses them, and the mirror-misplaced tests move to their subject's path. It is moves only: no test body, assertion or helper behaviour changes.

• 77 moves: 40 fixture assets and 37 C# files.
  • Helpers: each was classified at base 78d47a16 by its code consumers only. A helper stays in the root Support/ when more than one slice uses it.
  • Tests: the 14 tests mirror.txt listed move to their subject's path with Internal dropped, so mirror.txt goes from 14 rows to 0.
  • Self-tests: the four Support/Tests self-tests sit beside their subjects.
• Fixture identity is preserved. Every moved EmbeddedResource keeps its old LogicalName (OpenCode.Sdk.Tools.Tests.Fixtures.<old path with dots>) and is included exactly once. Copied None inputs (spec, profile, curation, SDK source) do not move.
• Forced edits only:
  • namespaces and using directives;
  • 27 OwnershipPlanData.Models(...) qualifications against the new Binding.Models namespace;
  • the repository-gate site lists (mirror.txt, collapsed-assertions.txt, file-size.txt);
  • the .csproj resource items;
  • the testing-style.md §1 slice sentence, which now covers SDK.Tests and Tools.Tests and names Tools' one deliberate root hold: the Roslyn compiler helpers, GeneratedSourceCompiler and its closure.
• Plan corrections, each decided by the code-consumer rule:
  • ModelPlacementFixture stays in the root Support/: its extension methods already had consumers in several slices.
  • RefreshScenarioData stays in the root Support/: a newer Commands consumer uses it.
  • TestRosterScenario and TrxContent move with the roster test to Commands/Support/. Their only code consumer is Commands/VerifyTestRosterCommandTests.cs, which this PR moves there; the plan had left them in TestRoster/Support/.

Part of TST-E7 (step 3b). Related: #403 records a cross-slice helper that predates this PR and is outside its inventory (Generator/Refresh/Support/RefreshScenario, also used by Commands).

🔄 Type of Change

• 🧪 Test improvements

🎯 Protocol & API Surface

• No change to the public API surface
• Nothing under external/, spec/ or src/ was touched

🧪 Testing

Clone baseline. The second commit carries only this PR's own tests-baseline delta: 23 using-header fingerprints added and 17 removed, re-paired because the moved files' using blocks now pair differently. There are no count changes, and no method-body pair is baselined.

The delta was derived from empty-set regenerations (quality-gates.md), because --update-baseline merges into the existing file:

• master 78d47a16 has 811 live fingerprints, and this head has 817;
• exactly their difference is applied to master's committed baseline;
• master's 28 stale entries are left for TST-E0c (verify-clones fails on a baseline entry no clone matches (stale allowance) #384).

Rebase. This branch was rebased onto #402's merge. The testing-style.md §1 conflict was merged hunk by hunk. Consumers of the moved helpers were re-checked on the new base: #398 and #404 added two Tools consumers (EmitterPlanFixture, EmitterSnapshot), both in their helper's own Generator/Emission slice, so no home changed.

Review. A fresh read-only review re-checked:

• helper ownership by code consumers, including Generator's GEN-E2 and GEN-E8 additions;
• the mirror moves and MirrorGuard;
• every moved resource's LogicalName;
• the baseline delta.

It raised two defects. The roster helpers are moved in this head. The §1 sentence's overclaim is fixed: the compiler-closure exception is named, and the pre-existing RefreshScenario case is #403.

✅ Completion Gate

Local check (quality-gates.md) on head 7c8753b8, base 78d47a16. Every step's own exit code was read:

• slopwatch: zero findings
• dotnet build --configuration Release: 0 warnings, 0 errors
• dotnet format whitespace / dotnet format style --severity warn: no changes
• verify-clones: passed, 0 new
• OpenCode.Sdk.Tools.Tests (all but NativeAot): 1794, 0 failed
• OpenCode.Sdk.Tests (all but ServiceElection/NativeAot): 8203, 0 failed. Its first attempt stopped at discovery because this worktree's pinned server was not yet installed; after bun install in external/opencode, it ran without a rebuild.

The run shared the host with a dispatched worker's read-only phase; no step needed a re-run because of it.

The host runs Bun 1.4.3, a maintainer-named off-pin host (OPENCODE_SDK_TESTS_BUN_MISMATCH=allow); the pin is 1.4.2.

✅ Code Quality Checklist

• No new suppression
• No tests skipped, weakened, or deleted to make the suite pass

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[Blind-Striker]

Correction to this PR's description: it says the baseline delta was derived from empty-set regenerations "because --update-baseline merges into the existing file". That reason is wrong. A direct probe shows --update-baseline rewrites each baseline to the current clones and drops entries no clone matches, as quality-gates.md says. The delta itself is correct: an empty-set regeneration yields the same live set.