Task: Bring ores.connections to the clean standard

ores.connections passes the Component Clean Standard for a component of kind All.

The inventory below shows that the component had no consumer left, so the task reached the standard by removing it. The item table records that outcome item by item. The story carries the decision.

• Every checklist item that applies to a component of kind All passes, or is recorded with the reason it does not apply.
• Every item that cannot pass is recorded as an exception with its reason.
• Every generator defect found is fixed at its source, with a regression test.
• No consumer is left without the component. The whole tree builds without it.

The original acceptance asked that ores.connections join COMPONENTS_UNDER_TEST. That cannot be met and is superseded. The component no longer exists.

Work the Component Clean Standard in its section order and record each item's result in the table below.

• Inventory, 2026-09-23. 0 entity, junction or operation models. 70 C++ files outside tests, 0 of them generated (0%). 0 messaging headers, 0 of them hand-written.
• Consumer census. The only reference in the tree was projects/CMakeLists.txt line 124, add_subdirectory(.../ores.connections). #include <ores.connections/...> appears nowhere outside the component. A grep for ores.connections, ores.connections/ and connection_manager across every .cpp, .hpp, CMakeLists.txt, *.cmake, *.ts and *.py outside the component returns nothing.
• The component is not a codegen component. It has no row in projects/ores.codegen/library/component_catalogue.org, and the codegen has no SQLite target. The generated entity, domain and repository templates hard-code the PostgreSQL shape: schema = "public", tenant, version and audit columns, and sqlgen::session on a connection pool. This component uses sqlgen::sqlite with a file path. So 66 of its files are generatable only after new generator capability, which a component with no consumer does not earn.
• 6 files are transitively dead inside the component as well: domain/recent_party.hpp, repository/recent_party_entity.hpp, repository/recent_party_mapper.{hpp,cpp} and repository/recent_party_repository.{hpp,cpp}. The only caller of that family, connection_manager::get_recent_parties, has no caller of its own. get_child_folders has no caller either, and resolve_connection has no production caller, only two cases in the service test.
• Stale reference sweep. projects/modeling/ (component_architecture.org, system_model_infrastructure.org, system_model_foundation.org, ores.puml), projects/ores.security/modeling/component_overview.org, projects/ores.compass/src/env_init.py, the .env-format log, and projects/ores.web/package-lock.json.
• Two sprint-history pages linked the deleted component's :ID:. The links are now plain text, so compass lint passes and the history still reads.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

No scenario applies. No user reaches the component, and a deletion has no behaviour to walk through. The build and the gates are the evidence.

ores.connections is removed, together with the references it left behind.

The clean standard is met by subtraction. H02 forbids dead code, and the component was dead by every measure: no C++ consumer, no web or shell role, no org model, and no codegen target.

Verification. The whole tree builds with compass build, exit 0, re-proved on the rebased tree. The ores.compass suite passes at 215 passed and 1 skipped, and the ores.codegen suite at 522. No test covers the env_init.py edit, because the removed block lay inside env_init.run and no test calls it, so that edit rests on the diff and the build. Every codegen gate passes, apart from the inherited repo-wide CMake source-list drift that G06 records. compass lint passes.

Verification rounds found defects, and every one is fixed. The first independent verifier read the first patch and found two, fixed in the follow-up commit. A Claude review found a stale backlog capture, fixed next. The rebase onto trunk surfaced the icon inventory entry. A second independent verifier confirmed the removal is sound and then failed the patch on statements in this record. Each finding is corrected here after I reproduced it. The last one matters most: no test covers the env_init.py edit, so that change rests on the diff and the build.

The removal is one commit and reverts on its own. The reference sweep is a second commit.