Task: Bring ores.orgmode to the clean standard

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

The inventory below shows that the component is a hand-written, model-less C++ library, so the model, protocol, wiring and shell sections do not apply. The item table records that outcome item by item. V06 records the one exception, and 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 broken. The whole tree builds with the component.

The original acceptance asked that ores.orgmode join COMPONENTS_UNDER_TEST. That cannot be met and is superseded, because ores.orgmode is not a codegen component. V06 records the reason.

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

• Inventory, 2026-09-26. 0 entity, junction or operation models. 14 C++ files outside tests, 0 of them generated (0%). 0 messaging headers, 0 of them hand-written.
• Consumer census. ores.shell consumes the library: orgmode_commands.cpp includes parser.hpp and resolver.hpp and ores.shell/application/src/CMakeLists.txt links ores.orgmode.lib. No other component references it. The library links only Boost and sqlite3.
• Complexity measure, from lizard over include/ and src/: 452 NLOC, 27 functions, average cyclomatic complexity 3.1. The top file is src/parser/parser.cpp, whose parse() is the single warning at CCN 21 over 90 NLOC. There is no generated source to skew the totals.
• The codegen's flat profile has never been used: no component_overview.org in the tree declares #+component_kind: flat. No component tree holds a generated stub file; the only occurrences of the stub template's text are the two *.mustache templates and their two archetype .org sources.
• Coverage, measured on 2026-09-26: 261 of the component's 271 executable lines in include/ and src/ are hit by its tests. The three uncovered groups are the defaulted move operations (not worth pinning), seven gcov artifacts on the static regex accessors, and normalise_table's empty guard, which its only caller makes unreachable. The measurement is an instrumented clang build with llvm-cov gcov; the repo's own coverage runs nightly through CTest.cmake and submits to CDash. The first measurement exposed three untested resolver error paths, now covered.
• The refreshed diagram is captured with build/scripts/generate_component_puml.py --project ores.orgmode and then cleaned up by hand. PlantUML resolves a relationship only when it sits in the same namespace block as its classes, so the hand-authored parser functions, compositions and resolver note live in the generated section, above the sentinel. A future capture run regenerates that section and the cleanup is redone.

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. The change touches a diagram, two generated build lists and two documents, and no user reaches the component. The build, the component suite and the gates are the evidence.

ores.orgmode is brought to the Component Clean Standard short of the registry entry, which the standard's own exception rule covers.

The component is a hand-written simple library, not a codegen component, so the model, protocol, wiring and shell sections do not apply. Each item that does not apply is recorded with its reason. The work that did apply: the two generated CMake source lists regained their generated marker, the component diagram was refreshed from the headers and re-rendered, the Foundation layer page gained the component, and the component overview gained its filetags.

Verification. The whole tree builds with compass build, exit 0. The ores.orgmode suite passes 62 assertions in 20 test cases, the ores.codegen suite 530, and the ores.compass suite 230 with 1 skipped. Coverage of the component's own sources, measured with an instrumented build, is 261 of 271 lines. Every codegen gate passes, apart from the inherited repo-wide CMake source-list drift that G06 records.

The one exception is V06. ores.orgmode does not join COMPONENTS_UNDER_TEST because it is not a codegen component. Neither available component kind fits it: the default flat scaffolds the stub trio, and composite rewrites the root CMakeLists.txt and stops the file lists generating. The proof is the two dry runs recorded in the table, which a reviewer can rerun.

Verification rounds found defects, and every one is fixed. An independent verifier reproduced the V06 dry run, the cmake check, the lizard totals, the three test-suite results and the consumer census, and confirmed the rendered image matches its source byte for byte. It then failed four statements in this record: the refuted Notes sentence, the contingent V06 framing, the unscoped H02 Qt claim and the missing join_paragraph_lines in the diagram. Each is corrected above after I reproduced it.

A Claude review of PR #2142 re-checked the build lists against the files on disk, the V06 catalogue state, the diagram fields and the Foundation layer entry, and found no blocking issue. It raised one real footgun, that a capture run drops the hand-authored arrows and note above the sentinel, so the warning now sits below the sentinel where a capture run preserves it. It raised one cosmetic table-alignment nit, dismissed above with its reason.

A second independent verification ran on the rebased head and returned PASS+NOTES. It reproduced the test counts, the diagram hash and the V06 dry run, and confirmed the three new resolver cases are behaviour tests that fail if the resolver misbehaves. Its one defect was that the rebase had made the recorded suite counts stale, corrected above. It could not reproduce the coverage line counts without a full instrumented rebuild, so it recorded that claim as unverified rather than disputed.