[gdluck]

Requested by A

Summary

Review round 2 (GreenGrassBlueOcean, 2026-11-09) applied in 4cdf1a8, a merge of master 32cca9e into the branch whose tree is byte-identical to the single rebased commit 9dc1e45 (a history rewrite was not possible from the session that pushed; a squash merge gives the same result, and the rebased commit can be force-pushed instead on request). git diff master is the same either way. The runtime change is argument validation on the WHO 2 builders plus one wire-hygiene fix; everything else is documentation and golden-corpus metadata. No decoded value, log line of a valid frame or accepted parser domain differs from master.

• OWNAutomationCommand: all seven builders (status, get_shutter_status, raise_shutter, lower_shutter, stop_shutter, set_shutter_level, report_shutter_level) pass their WHERE through master's require_address_where (fix(who1): crash guards for empty DIMENSION values, non-empty WHERE on WHO 1 / WHO 17 builders, DIMENSION 4 status_level / status_speed #155), as the WHO 1 builders do. A WHERE that is not a str / int, a bool, None, empty, blank, or containing *, ## or inner whitespace raises ValueError (leading / trailing whitespace is stripped). Before: raise_shutter(True) built *2*1*True## (log "Raising shutter None."), status(None) *#2*None##, raise_shutter("31##") *2*1*31####, raise_shutter("") *2*1*##, raise_shutter("*") the valid-looking *2*1**##, get_shutter_status("") the valid-looking *#2**10##, status("31*10") the DIMENSION 10 read *#2*31*10##. OWNCommand.parse stays permissive (*2*1*## still parses as an invalid OWNAutomationCommand).
• report_shutter_level: a string priority is kept verbatim ("001") only when it is ASCII digits. "００１" passed str.isdigit() and reached the wire as *#2*31*10*10*40*００１*0##; it is now encoded through int (*1*). OWNAutomationEvent is untouched (zero decode diff, as discussed in the review).
• WHO 2 documentation: the module docstring of OWNd/message/automation.py describes the WHO 2 advanced-command, DIMENSION 10 / 11 and WHERE grammar against the committed MyHOME captures: the LN4660M2 centralised control of a MyHomeServer1 plant putting *2*11#100#001#1*0## / *2*12#100#001#1*0## / *2*10#001#1*0## on the bus (issue_445); an F454 echoing Home Assistant's relative steps *2*11#STEP#001*31## / *2*12#STEP#001*31## (STEP 1, 2, 10) as *2*1000#11#STEP#001#1*31## / *2*1000#12#STEP#001#1*31## while the actuator reports *2*1*31## / *2*2*31##, DIMENSION 10 and its own *2*0*31## (issue_466, config_entry-myhome_F454_step_commands.json); an F461 2.0.11 ACKing *2*13*05## / *2*14*05## and translating them as *2*1000#13#001#1*05## / *2*1000#14#001#1*05## while the F411/2 at 05 emits nothing for any of the five advanced commands (f461_live_2026-11-04, test(f461): live F461 2.0.11 full debug log and bus sweep, 2026-11-04 (released OWNd 2.0.0b13) MyHOME#865). The un-replayed "NACKed by all twelve emulated gateways" claim is gone from the code, the tests and the CHANGELOG: neither *2*1*## nor *#2**10## is in own-firmware-oracle, and the guards are presented as argument validation only.
• WHO 9: AUX_WHAT_NAMES / AUX_WHAT_NAMES_UNVERIFIED (unused) are removed; the provenance caveat on the WHAT 2..10 names lives in the OWNAuxEvent and channel docstrings only. Log strings unchanged.
• Golden corpus: cover.dim.write.pos.75 (*#2*93*#10*75##) relabelled source: conflict (WHO_2.pdf / Encyclopedia who-2-automation/dimensions.md define no DIMENSION 10 write; no committed trace carries the frame; it pins parser shape only); cover.dim.write.tilt.30 renamed cover.dim.write.level.30.no_priority (DIMENSION 11 is absolute position; WHO 2 has no slat-angle dimension). README, docs/protocol_conformance_matrix.md, scripts/sync_documentation.py and tests/golden/README.md no longer say "tilt angles" / "Slat Tilt"; the golden README count is the actual 166 (tools/golden/validate_corpus.py: 11 files, 166 frames, PASS).

Review follow-up (round 2, 2026-11-09)

#	Item	Action
1	Un-replayed 12/12 firmware-oracle claim	Removed from CHANGELOG.md, automation.py (module and class docstrings) and tests/test_who2_audit_fixes.py; the docstring now says explicitly that neither frame has been replayed in own-firmware-oracle.
2	require_non_empty_where -> require_address_where	All seven builders switched; tests cover True, False, None, 12.5, b"31", ["31"], "31##", "31 1", "31\t1", "31\n1" (rejected) and " 0 ", "31\n" (stripped and accepted, as require_address_where does for WHO 1).
3	Dead AUX_WHAT_NAMES / AUX_WHAT_NAMES_UNVERIFIED	Deleted; caveat kept in the OWNAuxEvent / channel docstrings.
4	Docstring vs committed traces	Rewritten around issue_445, issue_466 and f461_live_2026-11-04 as described above; 13 / 14 noted as ACKed and translated by the F461 2.0.11.
5	Non-ASCII digits	report_shutter_level priority string requires .isascii(); OWNAutomationEvent left untouched (zero decode diff).
6	Merge state, title, body, scratch tags	Brought up to master 32cca9e (merge 4cdf1a8, tree = rebased commit 9dc1e45); title and body rewritten to the actual diff; test_a4_a5_* / test_a9_* / rows A4, A5, A9 / D56 labels removed; sync_documentation.py --check, verify_library_standards.py, validate_corpus.py (166) re-run.

On the follow-up question: agreed, no separate PR re-proposing OWNAutomationEvent advanced-WHAT decoding. The MyHOME f461_live_2026-11-04/README.md wording that conflates the F461's 1000# translations with F411/2 actuator support will be corrected in MyHOME, not here.

Testing

• ruff check OWNd tests setup.py scripts, mypy OWNd tests/typing, python scripts/verify_library_standards.py, python scripts/sync_documentation.py --check, python tools/golden/validate_corpus.py (166 frames, PASS).
• python -m pytest -q: 3194 passed; tests/test_who2_audit_fixes.py: 111 cases.
• MyHOME v2-phase1-architecture (6c50e358) test suite with this branch on PYTHONPATH, the way myhome-conformance runs it: see the follow-up comment for the result line.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[GreenGrassBlueOcean]

Thanks for auditing WHO 2, WHO 3, and WHO 9 and explicitly noting in the PR description that the Encyclopedia's WHO 2 captures are MyHOME's own captures rather than independent traces of the advanced commands.

Under our standing rule — if we don't have real, committed on-wire plant traces from physical hardware, we don't change behaviour — please trim this PR to the non-empty-WHERE builder guards and pure documentation fixes, and revert the doc-derived runtime changes.

What can stay

1. Non-empty WHERE guards on OWNAutomationCommand builders (A5): Rejecting empty where in up, down, stop, set_position, status, and advanced_status (*2*1*##, *#2**10## are NACKed 12/12 in the firmware oracle), while keeping OWNCommand.parse permissive.
2. Documentation & "Slat Tilt" cleanup: Removing the unverified "Slat Tilt" wording from README.md, docs/protocol_conformance_matrix.md, scripts/sync_documentation.py, and tests/golden/README.md, plus docstring notes on WHO 2 / WHO 9 specification caveats.

---

Blockers — Behaviour changes without committed plant traces (please revert)

#	Symbol / File	Behaviour change without real plant trace	Required action
1	OWNAutomationEvent (WHAT 10, 11, 12 & WHAT 13, 14) in OWNd/message/automation.py	Changes state (10/11/12 -> None) and is_closed (False -> None) on WHAT 10 / 11 / 12, changes WHAT 13 / 14 from state = 13/14 (is_opening / is_closing) to unknown_command_code (None), and adds speculative step, moves_to_end_point, priority_bitfield, priority_safety/high/medium, and selector decoders from WHO_2.pdf (ownkb:claim:c007501, c007502) with no independent physical advanced-shutter plant traces.	Revert OWNAutomationEvent runtime decoding and human_readable_log changes to master until independent on-wire captures of advanced WHO 2 shutter flows are committed.
2	OWNAutomationCommand.report_shutter_level & DIMENSION 10 decoding in OWNd/message/automation.py	Rejects state=15 and info outside {0, 12, 13, 14, 15} in report_shutter_level (forcing edits to test_expose_builders_infeasible_inputs in tests/test_message_protocol.py), drops DIMENSION 10 state outside 10..14, and appends info_name / alters 255 human_readable_log strings without trace evidence.	Revert report_shutter_level validation/log changes and DIMENSION 10 decoding changes to master.
3	cover.dim.write.pos.75 (*#2*93*#10*75##) in tests/golden/frames/who02_automation.yaml & tests/golden/corpus.json	Deletes cover.dim.write.pos.75 from the golden corpus.	Do not delete golden parser fixtures: keep the record in who02_automation.yaml / corpus.json and relabel it source: conflict with updated mcp_notes (the same pattern used in #157 for who18_energy.yaml).
4	_LOAD_GENERAL_STATUS in OWNd/message/base.py & OWNLoadCommand.general_status in OWNd/message/load.py (WHO 3)	Adds *#3*## special-casing in OWNMessage / OWNCommand and adds OWNLoadCommand.general_status() (*#3##) with zero physical F520 / F521 / F522 / F523 load-management plant captures (empty-bus gateway forwarding does not establish device replies).	Revert base.py and load.py runtime additions until we have a real WHO 3 load-management capture.
5	OWNAuxEvent.human_readable_log (WHO 9) in OWNd/message/auxiliary.py	Appends (WHAT N, unverified name) to human_readable_log for WHAT 2..10.	Revert the runtime human_readable_log change; keep the provenance caveat in AUX_WHAT_NAMES comments / OWNAuxEvent docstrings.

[gdluck]

Round 2 (review of 2026-11-09) applied in 4cdf1a8: a merge of master 32cca9e into the branch whose tree is byte-identical to the single rebased commit 9dc1e45 (git diff master is the same; a force-push was not possible from the session that pushed, so the history is not rewritten; a squash merge gives the same result, and I can force-push the rebased commit instead if you prefer). The description is rewritten to the actual diff.

#	Item	Action taken
1	Un-replayed 12/12 firmware-oracle claim	Removed from CHANGELOG.md, automation.py and the tests. Checked own-firmware-oracle main (*.cases, results) and tests/golden/firmware_oracle.json on master: no WHO 2 empty-WHERE frame anywhere; the docstring now says so and presents the guards as argument validation.
2	require_address_where on all seven builders	Done, including set_shutter_level and report_shutter_level. Rejected: True, False, None, 12.5, b"31", ["31"], "31##", "31 1", "31\t1", "31\n1". "31\n" strips to 31 and is accepted, exactly as require_address_where behaves for WHO 1 (" 12 "), so it is pinned as accepted rather than rejected.
3	Dead WHO 9 globals	AUX_WHAT_NAMES / AUX_WHAT_NAMES_UNVERIFIED deleted; provenance caveat kept in the OWNAuxEvent / channel docstrings.
4	Docstring vs committed traces	Rewritten around issue_445 (LN4660M2 centralised *2*11#100#001#1*0## / *2*12#100#001#1*0## / *2*10#001#1*0##), issue_466 (F454 relative steps *2*11#STEP#001*31## / *2*12#STEP#001*31##, STEP 1 / 2 / 10, 1000# echoes, actuator *2*1*31## / *2*2*31## / DIMENSION 10 / *2*0*31##) and f461_live_2026-11-04 (F461 2.0.11 ACKs *2*13*05## / *2*14*05##, translates *2*1000#13#001#1*05## / *2*1000#14#001#1*05##; the F411/2 at 05 emits nothing for the five advanced commands). One detail I could not confirm in the committed log: it holds no *2*15*05## frame, so the NACK on WHAT 15 is not cited.
5	Non-ASCII digits	report_shutter_level keeps a string priority only when .isascii() and .isdigit() ("００１" -> *1*, pinned). OWNAutomationEvent untouched: zero decode diff.
6	Merge state / title / body / scratch tags	Up to date with master (merge 4cdf1a8, tree = rebased 9dc1e45), mergeable; title and body rewritten; test_a4_a5_*, test_a9_*, rows A4, A5, A9, D56 removed; sync_documentation.py --check, verify_library_standards.py, validate_corpus.py (11 files, 166 frames) green; golden README count 166 is now asserted by a test.

Checks: ruff, mypy OWNd tests/typing, pytest 3194 passed (111 in tests/test_who2_audit_fixes.py); MyHOME v2-phase1-architecture 6c50e358 suite with this branch on PYTHONPATH: 5565 passed, 1 deselected (the known test_check_all_documentation_real_tree pin signal).

Agreed on the follow-up: no separate PR for OWNAutomationEvent advanced-WHAT decoding. The f461_live_2026-11-04/README.md wording in MyHOME (gateway 1000# translation vs F411/2 support) will be fixed in MyHOME.

---

Round 1 (2026-11-04), applied in `fe3c463`

Blocker	Action taken
1. OWNAutomationEvent WHAT 10 / 11 / 12, 13 / 14 and DIMENSION 11	automation.py rebuilt from master: state, is_closed, is_opening / is_closing as before, no unknown_command_code; step, moves_to_end_point, priority_bitfield, priority_safety / high / medium, selector, info_name, is_command_not_executed, target_level, is_go_to_level and their logs removed (the same DIMENSION 11 decoding / log change on OWNAutomationCommand.__init__ reverted too, since a log string on a valid frame is runtime behaviour).
2. report_shutter_level / DIMENSION 10	master restored (state 10..15, priority / info 0..255, master's 255 log, state outside 10..14 decoded as before); test_message_protocol.py restored verbatim.
3. cover.dim.write.pos.75	Restored at its original position, relabelled source: conflict with mcp_notes explaining that WHO_2.pdf / the Encyclopedia define DIMENSION 10 as request / response / event only and that no committed trace carries the frame.
4. WHO 3	base.py and load.py restored from master; tests pruned.
5. WHO 9	master's human_readable_log chain restored verbatim; the (WHAT N, unverified name) suffix and the state_name / is_state_name_verified properties removed.

[GreenGrassBlueOcean]

Checked fc34e2c against current master (32cca9e6), own-firmware-oracle, the committed MyHOME traces (issue_445, issue_466, and MyHOME#865), and downstream MyHOME consumers (cover.py, expose.py, gateway_events.py).

Does this PR solve a real problem?

• Runtime (OWNAutomationCommand WHERE guards): Marginal API hygiene only — fixes no bug or crash. On master, status("") (*#2*##), raise_shutter("") (*2*1*##), lower_shutter("") (*2*2*##), and stop_shutter("") (*2*0*##) already parse with is_valid is False and are blocked before touching the socket or UART by OWNCommandSession.send (OWNd/connection/command_session.py:283) and AsyncSerialTransport.send (OWNd/transport/serial.py:173) with OWNProtocolError(..., reason="invalid_frame"). The only is_valid is True frames previously reachable from those builders were get_shutter_status("") (*#2**10##, accepted by base.py only because WHO 13 uses *#13**0##), where="*" (*2*1**##, accepted only because WHO 5 uses *), and delimiter injection (status("31*10") -> *#2*31*10##). Every MyHOME caller (cover.py, expose.py) passes a pre-validated WHERE, and unlike WHO 1 *#1*## (MH202 crash, #155) or WHO 4 Dimension 22 (bt_termo crash, #156), there is no firmware crash vector here. Fail-early builder validation is fine for parity with #155, provided it uses master's require_address_where rather than the superseded require_non_empty_where.
• Documentation & golden metadata (WHO 2): Yes (errata cleanup). OpenWebNet WHO 2 has no wire dimension for slat tilt (MyHOME's Venetian blind tilt in cover.py is purely client-side timed motor-pulse estimation over standard UP / DOWN / STOP), and WHO_2.pdf defines no Dimension 10 write (*#2*93*#10*75##). Relabelling cover.dim.write.pos.75 as source: conflict, renaming/clarifying cover.dim.write.tilt.30, and removing "Slat Tilt" / "tilt angles" from WHO 2 wire-protocol docs fixes real documentation drift.
• WHO 9 module globals (AUX_WHAT_NAMES, AUX_WHAT_NAMES_UNVERIFIED): No. Dead unexported globals referenced nowhere in OWNd or MyHOME.

---

Summary of Required Changes

#	Severity	Location	Issue	Required Action
1	Blocker	CHANGELOG.md:8, OWNd/message/automation.py:24-25, 177-178, tests/test_who2_audit_fixes.py:4-5, 76	States 6× as empirical fact that *2*1*## and *#2**10## "are NACKed by all twelve emulated gateways in the firmware oracle" — neither frame exists in own-firmware-oracle or tests/golden/firmware_oracle.json.	Remove the unverified 12/12 firmware-oracle claim everywhere (cite WHO_2.pdf / Encyclopedia who-2-automation/addressing.md#where-forms), or actually replay *2*1*##, *#2*##, and *#2**10## in own-firmware-oracle first and cite the real suite.
2	High	OWNd/message/automation.py:38, 185-228, 264	Uses require_non_empty_where instead of master's require_address_where (#155): raise_shutter(True) passes mypy --strict (bool <: int) and builds invalid *2*1*True## (where=None, log "Raising shutter None."); status(None) -> *#2*None##; raise_shutter("31##") -> *2*1*31####; raise_shutter("31 1") -> *2*1*31 1##.	Switch all seven OWNAutomationCommand builders (set_shutter_level and report_shutter_level included) to require_address_where, and test bool, None, "31##", "31 1", and "31\n".
3	Medium	OWNd/message/scenario.py:75-94	AUX_WHAT_NAMES and AUX_WHAT_NAMES_UNVERIFIED are dead globals (not in OWNd/message/__init__.py __all__, not used by OWNAuxEvent.__init__, not imported anywhere).	Delete both dicts; keep the provenance caveat in the OWNAuxEvent / channel docstrings only.
4	Medium	OWNd/message/automation.py:4-5, 11-12, tests/test_who2_audit_fixes.py:8	Claims "no committed on-wire plant trace exercises the advanced forms" (contradicted by MyHOME#445 and MyHOME#466) and states unqualifiedly that 13 / 14 are "only DIMENSION 10 state values" (contradicted by F461 2.0.11 translating *2*13*05## / *2*14*05## in MyHOME#865).	Reword both docstring notes to match the actual committed plant captures.
5	Medium	OWNd/message/automation.py:67-95, 280-286	Non-ASCII digit leak vs #155 / #156: OWNAutomationEvent (DIMENSION 10) decodes fullwidth Unicode digits (*#2*21*10*１２*４０*００１*１４## -> state=12, current_position=40), and report_shutter_level("31", 40, 10, "００１", 0) emits non-ASCII wire bytes because .isdigit() lacks .isascii().	At minimum harden report_shutter_level's string priority check with .isascii() (and either guard DIMENSION 10 with .isascii() and .isdigit() for #155/#156 parity or leave OWNAutomationEvent untouched if keeping zero decode diff here).
6	Blocker	Branch merge state + PR Title / Body + tests/test_who2_audit_fixes.py	Branch is 30 commits behind master (mergeable_state: "dirty", conflicts in 6 files); PR title, summary, changes list, state diagram, and test count ("79 cases") still describe the reverted b931362 design; test file retains private scratch tags (rows A4, A5, A9, D56, test_a4_a5_*).	Rebase onto master, rewrite PR title and body to match the actual diff, re-run sync_documentation.py / corpus validation (verify fixture count in tests/golden/README.md), and strip scratch audit labels from tests/test_who2_audit_fixes.py.

---

Detailed Notes & Answer on the Follow-Up Proposal (#6633949587 / MyHOME#865)

1. Unverified firmware-oracle citation (*2*1*##, *#2**10##):

   • Checked own-firmware-oracle/suites/*.cases (ownd-audit-frames.cases only tested WHO 1 *#1*##, WHO 3, WHO 17, WHO 18, WHO 22; trace-corpus.cases only tested populated WHO 2 WHERE values), own-firmware-oracle/results/**/*.tsv, and tests/golden/firmware_oracle.json: zero WHO 2 empty-WHERE frames have been replayed. As your original PR body noted ("The WHO 2 rows are decode-side"), the 12/12 parenthetical from Round 1 review text was never replayed in the oracle and must not be recorded as an oracle verdict in docstrings, tests, or CHANGELOG.md.

2. Corrections on the cited MyHOME traces (issue_445, issue_466, and MyHOME#865):

   • issue_445 vs issue_466: issue_445 (MyHomeServer1 + LN-4660M2) carries centralized General WHERE=0 frames (*2*11#100#001#1*0##, *2*12#100#001#1*0##, *2*10#001#1*0##), not *2*11#1#001*31## / *2*12#1#001*31##. The WHERE=31 step frames (10%, 2%, 1%) are HA-originated commands in issue_466/config_entry-myhome_F454_step_commands.json (which F454 echoed as *2*1000#12#10#001#1*31##, while the actuator at 31 reported standard *2*2*31##, *2*0*31##, and Dimension 10 *#2*31*10*12*40*001*0## -> *#2*31*10*10*30*001*0##).
   • MyHOME#865 (f461_live_2026-11-04/F461_full_debug_log_2026-11-04T15-45-56.txt): Address 05 is a basic timed BT-F411/2 relay shutter, not an advanced actuator. Compare lines 583–596 (*2*2*05## -> gateway *2*1000#0*05##, actuator *2*0*05##, gateway *2*1000#2*05##, actuator *2*2*05##) against lines 607–650 (*2*11#1#001*05##, *2*12#1#001*05##, *2*10#001#1*05##, *2*13*05##, *2*14*05##): on all five advanced commands, only the F461 gateway's bt_autom daemon emitted *2*1000#...*05## command translations — the F411/2 at 05 emitted zero status frames and ignored every one of them. (Please also correct f461_live_2026-11-04/README.md:56 in MyHOME so gateway command translation is not conflated with F411/2 actuator support.)
   • WHAT 13 and WHAT 14 on F461 2.0.11: Lines 637–650 of that same log show F461 2.0.11 ACKing *2*13*05## and *2*14*05## and translating them onto the monitor session as *2*1000#13#001#1*05## and *2*1000#14#001#1*05## (while NACKing *2*15*05##). That confirms bt_autom treats WHAT 13 and WHAT 14 as valid command WHATs despite WHO_2.pdf listing 13 / 14 only under Dimension 10 states — directly illustrating why failing closed on 13 / 14 from the PDF alone would have broken real gateway traffic.

3. Should a separate PR re-propose OWNAutomationEvent advanced-WHAT decoding?

   • No. Downstream MyHOME (cover.py and gateway_events.py) already handles issue_445's *2*10/11/12#...*0## frames through is_opening / is_closing, already accounts for effective_what in (1, 2, 11, 12, 13, 14) on 1000# translations (cover.py:2576), and deliberately ignores Dimension 11 frames (cover.py:2582-2590) because shutter position comes from the actuator's Dimension 10 reports rather than command echoes. Pinning the existing issue_445 / issue_466 frames as community-plant-capture golden records against master's current decoding is fine if desired, whereas adding unused properties (step, moves_to_end_point, priority_bitfield, selector) or changing state to None solves no real problem.

[GreenGrassBlueOcean]

Neither *2*1*## nor *#2**10## has been replayed in own-firmware-oracle or tests/golden/firmware_oracle.json (ownd-audit-frames.cases only tested WHO 1 *#1*##). Drop the all twelve emulated gateways in the firmware oracle claim here and in automation.py / test_who2_audit_fixes.py unless those frames are actually added to own-firmware-oracle and replayed, and update this bullet to reflect require_address_where across all seven OWNAutomationCommand builders.

[GreenGrassBlueOcean]

Two factual fixes needed in this docstring block:

1. Line 5 ("no committed on-wire plant trace exercises the advanced forms"): MyHOME has committed traces exercising *2*11#100#001#1*0##, *2*12#100#001#1*0##, *2*10#001#1*0## (issue_445, MyHomeServer1 + LN-4660M2) and *2*11#10#001*31## / *2*12#10#001*31## / *2*12#2#001*31## / *2*12#1#001*31## (issue_466, F454 + 1000# echoes + Dimension 10 replies). Reword to reflect what is vs isn't captured on-wire.
2. Lines 11–12 (13 / 14 "listed only as DIMENSION 10 state values") & lines 24–25: note that while WHO_2.pdf lists 13 / 14 only under Dimension 10, physical F461 2.0.11 (bt_autom) ACKs *2*13*W## / *2*14*W## and translates them as *2*1000#13#001#1*W## / *2*1000#14#001#1*W## (MyHOME#865), and remove the un-replayed firmware oracle NACKs *2*1*## / *#2**10## claim on lines 24–25.

[GreenGrassBlueOcean]

After rebasing onto master, use require_address_where(where) (added in #155) instead of require_non_empty_where(where) across all seven OWNAutomationCommand builders (status, get_shutter_status, raise_shutter, lower_shutter, stop_shutter, set_shutter_level, and report_shutter_level). require_non_empty_where still lets bool (raise_shutter(True) passes mypy --strict and builds invalid *2*1*True##), None, "31##", and "31 1" slip through. Also drop the un-replayed firmware-oracle claim in the class docstring above (lines 177–178).

[GreenGrassBlueOcean]

Neither AUX_WHAT_NAMES nor AUX_WHAT_NAMES_UNVERIFIED is exported in OWNd/message/__init__.py (__all__), referenced by OWNAuxEvent.__init__ below, or used anywhere else. Please remove both unused module globals and keep the provenance caveat in the OWNAuxEvent / channel docstrings only.

[GreenGrassBlueOcean]

Strip the private review-session scratch tags (review-session rows A4, A5, A9, # ── A4 / A5:, (gate, D56), test_a4_a5_*, test_a9_*) and the un-replayed NACKed 12/12 in the firmware oracle comment here and on line 76, and expand the rejected-WHERE parametrization to cover True, False, None, "31##", "31 1", and "31\n" once switched to require_address_where.