refactor(worker): remove MessageParser (replaced by StreamAnalyzer)

tests/ClaudeDo.Worker.Tests/CLAUDE.md

@@ -0,0 +1,41 @@
+# ClaudeDo.Worker.Tests
+
+xUnit integration tests for the Worker and Data layers.
+
+## Framework
+
+- xUnit 2.5.3 with `xunit.runner.visualstudio`
+- No mocking library — custom sealed fakes (FakeClaudeProcess, FakeHubContext, FakeHubClients, FakeClientProxy)
+- Real SQLite databases per test via `DbFixture`
+- Real git repos for worktree tests via `GitRepoFixture`
+- coverlet for coverage collection
+
+## Test Infrastructure
+
+- **DbFixture** — creates unique temp SQLite DB, applies schema, cleans up DB + WAL/SHM files on dispose
+- **GitRepoFixture** — creates temp git repo with initial commit, configures user/email, handles Windows read-only .git cleanup. Tests skip if git is unavailable.
+
+## Test Areas
+
+| Area | File | What it covers |
+|------|------|----------------|
+| Repositories | `ListRepositoryTests` | CRUD, tag junctions |
+| | `TaskRepositoryTests` | CRUD, status transitions, agent tag filtering, effective tags, stale flip |
+| Runner | `WorktreeManagerTests` | Worktree creation, commit detection, error on non-git dir |
+| | `CommitMessageBuilderTests` | Slug generation, title/description truncation |
+| | `MessageParserTests` | NDJSON parsing, malformed input |
+| Services | `QueueServiceTests` | FIFO ordering, override slot contention, cancellation, active tracking |
+| | `StaleTaskRecoveryTests` | Flips orphaned running tasks to failed |
+
+## Conventions
+
+- Test classes implement `IDisposable` and create fixtures in constructor
+- Helper factory methods for entities: `MakeTask()`, `CreateListAsync()`, `SeedListWithAgentTag()`
+- Concurrency tests use `TaskCompletionSource` as gates for deterministic ordering
+- Git-dependent tests are conditionally skipped via `Skip = ...` when git is not available
+
+## Running
+
+```bash
+dotnet test tests/ClaudeDo.Worker.Tests
+```