[joelteply]

Under the new rule (no one edits a file over 1,000 lines), these eight files are each broken apart by responsibility. Every file now ends at most 1,000 lines. The moves are pure: test counts are identical before and after for each module.

File	Before → after	New files
inference/llamacpp_adapter.rs	1794 → 934	construction.rs 426, payload.rs 70, tests.rs 377
persona/response.rs	1504 → 918	sanitize.rs 200, tests.rs 392
commands/genome/teach.rs	1410 → 944	progress.rs 200, status.rs 95, tests.rs 185
modules/agent.rs	1393 → 841	tools.rs 558
cognition/generate_response.rs	1366 → 644	tests.rs 723
inference/airc_remote/transport.rs	1160 → 757	tests.rs 404
inference/handle_module.rs	1094 → 663	tests.rs 432
persona/rag_inspect.rs	1034 → 684	tests.rs 351

Beyond pure moves, two changes:

• teach.rs re-exports TeachProgress, subscribe_teach_progress and the status types, so external paths are unchanged.
• Two tracing::warn! calls that moved got an explicit target:, so their log target is unchanged.
• The moved tests.rs files carry #![cfg(test)], the repo's existing convention. That made layer_discipline skip test-module files, which is a rule change: please review it.

Baselines lowered to measured counts:

• lines over 1,000: 155,781 → 153,027
• unjustified unwraps: 2,123 → 2,113
• upward imports: 253 → 252
• underived bounds: 159 → 158

cargo check: 122 warnings before and after.

Each module's cargo test filter has the same counts before and after. Run on the M5; the full suite is left to CI.

🤖 Generated with Claude Code

[joelteply]

[P2] Restrict the test-file exemption to an actual file-level attribute (1ac909c)

In core/source-hygiene/src/layer_discipline.rs the new lines().any(trim_start().starts_with("#![cfg(test)]")) returns no violations for the entire file whenever any nested module contains this inner attribute. For example a production inference file with use crate::persona::home; followed by mod tests { and an indented #![cfg(test)] now skips its production upward import. A matching line inside a multiline comment/string has the same problem. This is a semantic change to enforcement, separate from the pure moves.

Recognize only the source file's leading inner-attribute region (with comments handled), or use a shared parser that proves file scope. Add fixtures for a genuine file-level test module and for a production import plus nested test-only module; the latter must still report the production violation. No build run on a serving node. Full inference split review remains pending; this is not an approval of the PR.

[joelteply]

Completed the owned inference slice source review at 1ac909c: airc_remote/transport, handle_module and llamacpp_adapter.

Compared old parent versus new parent plus children with a whitespace-normalized line multiset, excluding module/import/doc scaffolding and normalizing pub(super). Transport: 12 tests before/after, no body-line difference. Handle module:16/16, no body-line difference. Llama adapter:13/13; only additions are two explicit tracing targets preserving the original module target. Payload construction boundaries and parent imports inspected. This supports source-level equivalence for these three cuts; it is not compilation or runtime proof and does not cover the other directories.

The layer_discipline exemption finding above remains unresolved at this head, so this is not whole-PR approval. CI was not reported when checked; no build run on a serving node.

[joelteply]

Scoped source PASS at b3ddc45 for the newly moved inference/handle_module and inference/airc_remote/transport tests. Compared original inline test bodies immediately before d9f0f8d and 8c63901 against current child files: exact match after four-space module dedentation and removing child cfg marker. Both parent cfg(test) module declarations and child #![cfg(test)] markers remain; module paths are unchanged. Counted16/12 ordinary test annotations in these files. Production portions of those commits are untouched. Current reported CI is green; no local build run. This is only the two-file extraction slice, not approval of inherited hygiene/other owners' work.

[joelteply]

Superseded for landing by #5228, which contains this PR's commits in full (merged down, not rebased) and passes the new oversized-edits gate. On its own, this PR now fails that gate: it creates or edits files over 1,000 lines that only #5228 splits (recall_faculty/tests.rs 1,245, training_trigger/submit/tests.rs 1,332, forge.rs 1,002). Landing order: #5227, then #5228. Converted to draft, not closed, so the scoped reviews stay readable.