[trissim]

Summary: 3 phase greps 1/0/0 to 0/0/0. 14 annotations retyped in 7 modules (1 the phase pattern sees, 13 it cannot). 4 constructor re-checks deleted. 22 lines added, 33 removed. No wire, schema or stored-row change.

Plan phase: id128_nominal/root_ids (RootIdsPhase). Earlier nominal PRs had already cleared nearly all of the plan's 127 sites, so the phase's own remainder on main was one line.

Grep counts (git grep -nE PATTERN -- src ':!src/agent_comms/skills/', base d6084cb)

Pattern	Before	After
\b(wire_root_id|root_id|lookup|recipient_lookup): (str\b|Id128\b|Annotated\[(str|Id128)\b)	1 (pi_rule_gate.py:102)	0
Id128Text\.decode\((self\.)?(owner\.)?(wire_root_id|root_id|lookup)\)	0	0
FieldCodec\.decode\(Annotated\[str, *Id128Text\]	0	0
Wider check, not a phase gate: any \w*(root_id|lookup)\w* annotated str, Id128, Annotated[str|Id128, Optional[str, list[str], tuple[str	14	0

What changed

Same kind under names the phase pattern cannot see, so retyped with it (one codemod, not committed):

Annotation	Producers
pi_rule_gate.py:102 typed_by(lookup)	pi_rule_gate.py:96 require_recorded().lookup, typed at registry_provenance.py:90
attempt_start.py:30 AttemptStart.owner_lookup	selected_turn.py:125-128 participant.lookup, typed at selected_participant.py:57
coordination_response.py:109 resolve_retained(owner_lookup)	coordination_response.py:266 (recipient_lookup, typed at :205, :231), attempt_recovery.py:305 (incarnation.lookup)
execution_store.py:41, :54 create, create_after_triage	selected_participant.py:94, selected_turn.py:263, selected_triage.py:137, all participant.lookup
native_input_record.py:203,207,344,358	native_runtime_input.py:222,261, from transcript_sources.py:642 self.recipient.lookup
native_input_record.py:452 NativeInputRecord.owner_lookup	Parent annotation; overridden with ThreadLookup by all three stored subclasses: native_runtime_input.py:277, historical_native_inputs.py:54, native_prompt_binding.py:58
bus.py:400,411, bus_tables.py:65,104 sender_lookups	goal_management.py:247, goal_waits.py:207, pi_rule_gate.py:106, all incarnation...lookup

Wire: nothing here changes what is encoded. Each retype is a parameter, or AttemptStart (never passed to FieldCodec), or the parent annotation every stored subclass overrides. No env, stored-file or older-runtime path produces these values. No installer wire record is touched, so older workers read the same bytes.

What was deleted

Each was a check on a field already declared ThreadLookup; a value that passed the decode is 32 hex, so non-empty and bounded said nothing further.

Deleted	Boundary where the value entered typed
CurrentExecutions.__post_init__ owner_lookup checks, and the three imports they left unused	Field coordination_tables/executions.py:231; written participant_store.py:76 from lookup: ThreadLookup
ExecutionRecord.__post_init__ owner_lookup loop	Field coordination_tables/executions.py:346; written execution_store.py:106 from the parameter typed here
AttemptRecord.__post_init__ owner_lookup entry	Field coordination_tables/attempts.py:183; written attempt_start.py:99 from ExecutionRecord.owner_lookup
OwnerFence.__post_init__ owner_lookup entry	Field owner_fence.py:33; sole constructor attempt_start.py:41, from AttemptStart.owner_lookup

Behaviour change, stated: a directly constructed row holding an empty or over-long plain str lookup no longer raises in __post_init__. Only tests construct these with raw text (test_s3_legality_behavior.py:222, test_optional_awareness_projection.py:324), and neither asserts that refusal. No table DDL changes.

Left alone, on purpose

• acp_extension.py:643 wire_root: str holds a filesystem path (thread_configuration.py:27 str(self.root.resolve())), not a RootId.
• wake_policy.py:93,122 FieldCodec.decode(ThreadLookup, lookup) are from_wire boundaries; the one decode stays.

Callers I could not prove

• native_runtime_input.py:210,229,257 and execution_store.py:67 take owner_lookup with no annotation. I traced their callers by hand; mypy does not check these parameters, and no grep will catch one going wrong.
• Mypy ran only on the 10 changed files with imports silenced, so call sites in other modules were read by hand, not machine-checked.

Checks run

• tests/test_s3_legality_behavior.py tests/test_coordination.py: 76 passed.
• ruff --select F,I: per-file counts equal main's on all 10 files.
• mypy on the 10 files: no new lookup error. The 3 lookup lines it reports (coordination_response.py:232 twice, :239) are unedited code.
• No local full suite and no local NRA gate; CI runs debt-src.

Overlap

Open PRs #3566, #3670, #3614, #3616 and #3607 touch bus.py, bus_tables.py or native_input_record.py. The plan's "no phase edits files of an open PR" dates from dfc7d8976 and named #2914. None of the diffs I could read edits these annotation lines (#3566, #3670, #3616, #3607); #3614's diff is too large for the API, so it is unchecked.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.