diff --git a/CHANGELOG.md b/CHANGELOG.md
index 20342404..30f396af 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@ Sections dated before 2026-09-19 predate the cycle and stay as they are.
 
 ## Unreleased
 
+- fix(process): **follow-ups to the change log (#306): baseline cost, CRLF registers, gate wording.** (1) `skills/change-intake.md` went baseline for lead+ba at Stages 2–7 at 1,804 words without paying for it, against `render-routing.sh`'s "a new baseline row is paid for elsewhere" rule. It is now a 331-word core (trigger plus its one-line failure case, row columns, class → entry artifact, size → ask, Owners — now the single source the runbook and `agent-roles.md` point to, with the inline-edit/no-`UNSYNCED` rule — and Done when); the why-table, the wrong-vs-right example, the full tables, apply/prove steps, the owners table and the gate-check explanation moved verbatim to new on-demand `skills/change-intake-detail.md` (lead, ba, architect; Stages 2–7). Baseline tier 79,997 → 78,524 words (`render-routing.sh --check`, `LC_ALL=C`). (2) A `PROJECT.md` saved with CRLF line endings hid the section: `mxtk_change_log_has_section` never matched `## Change log\r`, so gate-check warned "no section", `sync-project.sh` appended a second empty one, and the scan counted 0 rows. Both readers in `bin/lib/change-log.sh` now drop a trailing `\r`; new fixture case T9 in `tests/wave2/test-change-log.sh` (CRLF copy scans, gates and syncs identically to the LF one; 5 of its cases fail on the previous code). The fixture also drops `pipefail`, which let an early-matching `grep -q` SIGPIPE its `printf` and fail T7 intermittently. (3) The skill's "Done when" expected `0 of M … not yet applied`, a string gate-check never prints; it now names the `PASS` line gate-check actually prints — MendixMau
 - learn(testing-shape, test-result-audit): **three test-strength rules, each with a denominator: refusals probed on the server, every undeclared 4xx a failure, one clean-database final run** ([#172](https://github.com/MendixMau/mxcli-project-toolkit/issues/172)). A field comparison of two builds against one frozen spec found the toolkit build checked permission refusals only in the client (a hidden menu item), logged unexpected 400/422 responses as information, and reached 20/20 by stitching three runs. `testing-shape.md` §7 gains three report lines (`N of N restricted pairs probed server-side`, `undeclared 4xx/5xx: 0 of M requests — each fails its step`, `one clean-DB run <id>: N of N scenarios incl. timers, not stitched`); `test-result-audit.md` gains check 2e, which carries the reasons and the field evidence and asks for those numbers. Both baseline skills stay inside the 80,000-word baseline budget (79,997): the reasons live in the on-demand audit skill, and two redundant sentences in `testing-shape.md` §4's verdict paragraphs were folded (the empty-set corollary is stated in the Trace row, `harness-architecture.md` and `report-schema.md`). The backing inbox note is trimmed to its unpromoted "Keep" items. — MendixMau
 - learn(learned-detection-gaps, module-review): **two UI defects every model gate passes: a lazy-loaded combo box filter hides its preset, and a theme that pins row heights by position lets the next row cover an inserted widget** ([#299](https://github.com/MendixMau/mxcli-project-toolkit/issues/299)). Both passed check, mxbuild, lint and the scripted checks on a 127-module production app and were found only in the running app. Two register rows with the first oracle and the pre-build check (pluggable properties that change *when* data loads; grep the theme for fixed `height` on positional row classes before an insert), and two LOOK lines in `module-review.md` §4c with denominators (`N of N filters show their preset on first load`, `N of N widgets inserted into an existing container fully visible`). — jopterhorst
 - fix(app-restart, test-stack-up): **a port held by another process before `run --local` is named and reported as "instrument blocked" (exit 4), never killed** ([#262](https://github.com/MendixMau/mxcli-project-toolkit/issues/262)). Studio Pro had started its own runtime on the app/admin ports a restart needed; `mxcli run --local` then failed with "Authentication failed" (the admin API answering was Studio Pro's), the next journey ran against Studio Pro's runtime, and four more runs were NOT RUN. `app-restart.sh` now checks the app port and the admin port (record's `adminPort`, else app + 10) before stopping anything — a holder outside the recorded process tree exits 4 with its name, pid and parent — and again after the stop, where anything still holding a port is by definition not ours. `test-stack-up.sh` runs the same check wherever it would tell you to start `run --local` (`--no-docker`, no Docker daemon, `--check` with a leftover record), on the record's ports, `RUN_LOCAL_APP_PORT`/`RUN_LOCAL_ADMIN_PORT`, or mxcli's defaults 8080/8090. Holder lookup: lsof, then ss, then netstat + tasklist on Windows; unnamed holders still block when nothing of ours is running. Fixtures: `test-app-restart.sh` +2 cases (14 assertions, all fail on the old script), `test-stack-up-local.sh` T7 (4 of 6 fail on the old script). — AlvinnYeo
diff --git a/README.md b/README.md
index 50f41d55..299231b6 100644
--- a/README.md
+++ b/README.md
@@ -511,6 +511,7 @@ Every mxcli project has a `.ai-context/skills/` directory (bundled by `mxcli ini
 
 | Task | Skill to load |
 |---|---|
+| Applying or proving a change request (CR row), a change whose class, size or status is unclear, or a "Change log:" gate line to explain: full class/size/ask tables, the wrong-vs-right example, apply and prove steps, the owners table and what the warn-only gate check counts | `skills/change-intake-detail.md` |
 | Deep, adaptive interview on one topic, on demand, when a checkpoint's 2+1 or a single question batch isn't enough | `skills/grill-mode.md` |
 | CAC-1, closing Stage 0 in EVERY entry mode — scope IN: full scope or a slice, and in what order. Opens with a brainstorm, not options | `skills/checkpoints/checkpoint-scope.md` |
 | CAC-1b, after Stage 1 — scope OUT: is what extraction produced what you meant. No gate stops you; run it late against the BRDs if it was skipped | `skills/checkpoints/checkpoint-extraction.md` |
diff --git a/ROUTING.md b/ROUTING.md
index 2c55707c..a2c3d1da 100644
--- a/ROUTING.md
+++ b/ROUTING.md
@@ -38,6 +38,7 @@ picks the row up. That is the whole procedure — there is no second list to rem
 | Any pass whose input is missing, stale or unresolvable — before recording UNMEASURED, N/A or a silent skip: name what was missing, say what you assessed against instead, still deliver a verdict | `skills/degrade-to-judgement.md` | all | - | baseline |
 | Putting a question TO the user — any gate, any stage: ask in chat not in a file, two named options plus your recommendation, one batch per gate then end the turn | `skills/interview-protocol.md` | lead,ba,architect | P,0,1,2,3,4,5,6,7 | baseline |
 | The user gives feedback or asks for a change mid-project — "change X", "it also needs Y", a remark on a screenshot or the running app, a stakeholder's ask relayed in chat: log a CR row in PROJECT.md ## Change log BEFORE any edit, size it, then edit the highest artifact it alters (BRD, decision, architecture, wireframe), never the MDL first | `skills/change-intake.md` | lead,ba | 2,3,4,5,6,7 | baseline |
+| Applying or proving a change request (CR row), a change whose class, size or status is unclear, or a "Change log:" gate line to explain: full class/size/ask tables, the wrong-vs-right example, apply and prove steps, the owners table and what the warn-only gate check counts | `skills/change-intake-detail.md` | lead,ba,architect | 2,3,4,5,6,7 | ondemand |
 | Deep, adaptive interview on one topic, on demand, when a checkpoint's 2+1 or a single question batch isn't enough | `skills/grill-mode.md` | ba,architect | P,0,1,2,3,4,5,6,7 | ondemand |
 | Any stage transition — the 2+1 format every CAC uses, and the one-register rule (answers land in PROJECT.md, never in a separate state file). The seven CACs themselves are routed per stage in the situational table | `skills/checkpoints/checkpoint-template.md` | lead,ba,architect | 0,1,2,3,4,6,7 | baseline |
 | CAC-1, closing Stage 0 in EVERY entry mode — scope IN: full scope or a slice, and in what order. Opens with a brainstorm, not options | `skills/checkpoints/checkpoint-scope.md` | ba,architect | 0 | ondemand |
diff --git a/agents/architect-agent.md b/agents/architect-agent.md
index 5e6927a3..bada50f3 100644
--- a/agents/architect-agent.md
+++ b/agents/architect-agent.md
@@ -44,6 +44,7 @@ This table is your whole list. The baseline table in the project's CLAUDE.local.
 | `skills/tool-output-is-not-ground-truth.md` | Any time an exit code, a tool's output or a subagent's report is about to become a stated finding — verify before you conclude |
 | `bin/status.sh` | The first command of every session, and any time someone asks "where are we" or "what next" — one screen: stage, done/overdue, the ONE next action, from the instruments, never from memory |
 | `skills/retesting-learned-rules.md` | Before obeying any learned-* STOP or workaround that costs a detour — probe the binary you actually have, then stamp the verdict back into the rule |
+| `skills/change-intake-detail.md` | Applying or proving a change request (CR row), a change whose class, size or status is unclear, or a "Change log:" gate line to explain: full class/size/ask tables, the wrong-vs-right example, apply and prove steps, the owners table and what the warn-only gate check counts |
 | `skills/grill-mode.md` | Deep, adaptive interview on one topic, on demand, when a checkpoint's 2+1 or a single question batch isn't enough |
 | `skills/checkpoints/checkpoint-scope.md` | CAC-1, closing Stage 0 in EVERY entry mode |
 | `skills/checkpoints/checkpoint-architecture.md` | CAC-3, after BRD validation and before architecture locks |
diff --git a/agents/ba-agent.md b/agents/ba-agent.md
index 92ef81ee..67383caf 100644
--- a/agents/ba-agent.md
+++ b/agents/ba-agent.md
@@ -56,6 +56,7 @@ This table is your whole list. The baseline table in the project's CLAUDE.local.
 | `skills/tool-output-is-not-ground-truth.md` | Any time an exit code, a tool's output or a subagent's report is about to become a stated finding — verify before you conclude |
 | `bin/status.sh` | The first command of every session, and any time someone asks "where are we" or "what next" — one screen: stage, done/overdue, the ONE next action, from the instruments, never from memory |
 | `skills/retesting-learned-rules.md` | Before obeying any learned-* STOP or workaround that costs a detour — probe the binary you actually have, then stamp the verdict back into the rule |
+| `skills/change-intake-detail.md` | Applying or proving a change request (CR row), a change whose class, size or status is unclear, or a "Change log:" gate line to explain: full class/size/ask tables, the wrong-vs-right example, apply and prove steps, the owners table and what the warn-only gate check counts |
 | `skills/grill-mode.md` | Deep, adaptive interview on one topic, on demand, when a checkpoint's 2+1 or a single question batch isn't enough |
 | `skills/checkpoints/checkpoint-scope.md` | CAC-1, closing Stage 0 in EVERY entry mode |
 | `skills/checkpoints/checkpoint-extraction.md` | CAC-1b, after Stage 1 |
diff --git a/bin/gate-check.sh b/bin/gate-check.sh
index d24a021c..3c7e26e1 100755
--- a/bin/gate-check.sh
+++ b/bin/gate-check.sh
@@ -1768,12 +1768,12 @@ stage_protocol_paths() {
     P)  echo "skills/interview-protocol.md skills/grill-mode.md skills/agent-roles.md skills/bootstrap-project.md skills/cloud-dev-environment.md skills/existing-app-change.md skills/app-analysis.md skills/mendix-epics-api.md skills/corpus-extraction-integrity.md skills/platform-link.md" ;;
     0)  echo "skills/interview-protocol.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-scope.md skills/source-triage.md skills/image-transcription.md skills/small-project-tier.md skills/existing-app-change.md skills/app-analysis.md skills/layering-review.md skills/module-dependency-review.md skills/microflow-loop-antipatterns.md skills/assess-migration.md skills/migration-pipeline.md skills/migrate-general.md skills/migrate-outsystems.md skills/source-os11.md skills/os-xml-schema.md skills/source-node-express-react.md skills/document-discovery.md skills/extractor-quality-loop.md skills/qa-loop-goal-pattern.md skills/mendix-epics-api.md skills/corpus-extraction-integrity.md skills/gate-check-file-locations.md" ;;
     1)  echo "skills/interview-protocol.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-extraction.md skills/agent-roles.md skills/image-transcription.md skills/small-project-tier.md skills/migration-pipeline.md skills/source-os11.md skills/os-xml-schema.md skills/source-node-express-react.md skills/document-discovery.md skills/extractor-quality-loop.md skills/kb-generation.md skills/corpus-extraction-integrity.md" ;;
-    2)  echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-brd.md skills/checkpoints/checkpoint-architecture.md skills/agent-roles.md skills/image-transcription.md skills/small-project-tier.md skills/kb-generation.md skills/brd-generation.md skills/brd-validation.md" ;;
-    3)  echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-design.md skills/agent-roles.md skills/small-project-tier.md skills/mendix-best-practices-index.md skills/layering-review.md skills/architecture-blueprint.md skills/modularize-domain.md skills/design-artifacts.md skills/brd-to-build-plan.md skills/workflow-structure-rules.md skills/learned-mdl-cannot-express.md" ;;
-    4)  echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-build.md skills/agent-roles.md skills/small-project-tier.md skills/module-brief.md skills/module-folder-convention.md skills/brd-to-build-plan.md skills/coverage-ledger.md skills/workflow-structure-rules.md skills/rest-integration-first-time-right.md skills/learned-constants-and-secrets.md" ;;
-    5|build-ready) echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/agent-roles.md skills/module-brief.md skills/learned-mdl-preflight.md skills/module-folder-convention.md skills/learned-microflow-patterns.md skills/microflow-preflight.md skills/mendix-best-practices-index.md skills/ui-preflight-pages.md skills/design-spacing.md skills/learned-stylegallery.md skills/ui-loop.md skills/module-review.md skills/testing-shape.md skills/microflow-loop-antipatterns.md skills/iterative-build-loop.md skills/mdl-cookbook-microflows.md skills/build/mdl/oneshot-mdl-method.md skills/learned-page-patterns.md skills/oneshot-page-structure-patterns.md skills/mendix-agents.md skills/mendix-agent-ui.md skills/mendix-agent-setup.md skills/fixture-seeding.md skills/journey-proof.md skills/unhappy-path-testing.md skills/monkey-test.md skills/e2e-locators-that-lie.md skills/report-schema.md skills/harness-architecture.md skills/process-coherence-pass.md skills/lint-that-actually-runs.md skills/improvement-register.md skills/journey-examples.md skills/wiring-sweep.md skills/learned-workflow-patterns.md skills/workflow-structure-rules.md skills/rest-integration-first-time-right.md skills/bug-submission-checklist.md skills/empty-widget-triage.md skills/learned-sidebar-collapse-icons.md skills/learned-popup-navigation.md skills/learned-datagrid-customcontent-binding.md skills/learned-popup-feedback-pattern.md skills/learned-mdl-cannot-express.md skills/learned-css-that-never-applied.md skills/learned-detection-gaps.md skills/learned-dg2-patterns.md skills/learned-file-upload-widget.md skills/security-is-not-a-later-script.md skills/learned-local-db-confusion.md skills/full-harness-audit.md skills/test-result-audit.md skills/finding-disposition.md skills/preview-over-hub-tunnel.md skills/walking-skeleton.md skills/platform-link.md skills/teamserver-alignment.md skills/learned-constants-and-secrets.md" ;;
-    6)  echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-cutover.md skills/agent-roles.md skills/mendix-best-practices-index.md skills/module-review.md skills/testing-shape.md skills/existing-app-assurance.md skills/app-analysis.md skills/module-dependency-review.md skills/microflow-loop-antipatterns.md skills/qa-loop-goal-pattern.md skills/mendix-agent-setup.md skills/e2e-harness-base.md skills/learned-db-assertions.md skills/fixture-seeding.md skills/journey-proof.md skills/unhappy-path-testing.md skills/monkey-test.md skills/e2e-locators-that-lie.md skills/learned-skill-ux-audit.md skills/learned-skill-scope-delta.md skills/report-schema.md skills/harness-architecture.md skills/process-coherence-pass.md skills/e2e-evidence-report.md skills/record-demo-video.md skills/share-demo-package.md skills/lint-that-actually-runs.md skills/improvement-register.md skills/journey-examples.md skills/wiring-sweep.md skills/workflow-structure-rules.md skills/bug-submission-checklist.md skills/empty-widget-triage.md skills/anonymize-client-app-for-demo.md skills/learned-css-that-never-applied.md skills/learned-detection-gaps.md skills/learned-local-db-confusion.md skills/full-harness-audit.md skills/test-result-audit.md skills/finding-disposition.md skills/handoff-to-studio-pro.md skills/preview-over-hub-tunnel.md skills/platform-link.md skills/teamserver-alignment.md skills/learned-constants-and-secrets.md" ;;
-    7)  echo "skills/interview-protocol.md skills/change-intake.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-cutover.md skills/close-the-loop.md skills/share-demo-package.md skills/handoff-to-studio-pro.md skills/platform-link.md skills/teamserver-alignment.md skills/deploy-to-sandbox.md" ;;
+    2)  echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-brd.md skills/checkpoints/checkpoint-architecture.md skills/agent-roles.md skills/image-transcription.md skills/small-project-tier.md skills/kb-generation.md skills/brd-generation.md skills/brd-validation.md" ;;
+    3)  echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-design.md skills/agent-roles.md skills/small-project-tier.md skills/mendix-best-practices-index.md skills/layering-review.md skills/architecture-blueprint.md skills/modularize-domain.md skills/design-artifacts.md skills/brd-to-build-plan.md skills/workflow-structure-rules.md skills/learned-mdl-cannot-express.md" ;;
+    4)  echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-build.md skills/agent-roles.md skills/small-project-tier.md skills/module-brief.md skills/module-folder-convention.md skills/brd-to-build-plan.md skills/coverage-ledger.md skills/workflow-structure-rules.md skills/rest-integration-first-time-right.md skills/learned-constants-and-secrets.md" ;;
+    5|build-ready) echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/agent-roles.md skills/module-brief.md skills/learned-mdl-preflight.md skills/module-folder-convention.md skills/learned-microflow-patterns.md skills/microflow-preflight.md skills/mendix-best-practices-index.md skills/ui-preflight-pages.md skills/design-spacing.md skills/learned-stylegallery.md skills/ui-loop.md skills/module-review.md skills/testing-shape.md skills/microflow-loop-antipatterns.md skills/iterative-build-loop.md skills/mdl-cookbook-microflows.md skills/build/mdl/oneshot-mdl-method.md skills/learned-page-patterns.md skills/oneshot-page-structure-patterns.md skills/mendix-agents.md skills/mendix-agent-ui.md skills/mendix-agent-setup.md skills/fixture-seeding.md skills/journey-proof.md skills/unhappy-path-testing.md skills/monkey-test.md skills/e2e-locators-that-lie.md skills/report-schema.md skills/harness-architecture.md skills/process-coherence-pass.md skills/lint-that-actually-runs.md skills/improvement-register.md skills/journey-examples.md skills/wiring-sweep.md skills/learned-workflow-patterns.md skills/workflow-structure-rules.md skills/rest-integration-first-time-right.md skills/bug-submission-checklist.md skills/empty-widget-triage.md skills/learned-sidebar-collapse-icons.md skills/learned-popup-navigation.md skills/learned-datagrid-customcontent-binding.md skills/learned-popup-feedback-pattern.md skills/learned-mdl-cannot-express.md skills/learned-css-that-never-applied.md skills/learned-detection-gaps.md skills/learned-dg2-patterns.md skills/learned-file-upload-widget.md skills/security-is-not-a-later-script.md skills/learned-local-db-confusion.md skills/full-harness-audit.md skills/test-result-audit.md skills/finding-disposition.md skills/preview-over-hub-tunnel.md skills/walking-skeleton.md skills/platform-link.md skills/teamserver-alignment.md skills/learned-constants-and-secrets.md" ;;
+    6)  echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-cutover.md skills/agent-roles.md skills/mendix-best-practices-index.md skills/module-review.md skills/testing-shape.md skills/existing-app-assurance.md skills/app-analysis.md skills/module-dependency-review.md skills/microflow-loop-antipatterns.md skills/qa-loop-goal-pattern.md skills/mendix-agent-setup.md skills/e2e-harness-base.md skills/learned-db-assertions.md skills/fixture-seeding.md skills/journey-proof.md skills/unhappy-path-testing.md skills/monkey-test.md skills/e2e-locators-that-lie.md skills/learned-skill-ux-audit.md skills/learned-skill-scope-delta.md skills/report-schema.md skills/harness-architecture.md skills/process-coherence-pass.md skills/e2e-evidence-report.md skills/record-demo-video.md skills/share-demo-package.md skills/lint-that-actually-runs.md skills/improvement-register.md skills/journey-examples.md skills/wiring-sweep.md skills/workflow-structure-rules.md skills/bug-submission-checklist.md skills/empty-widget-triage.md skills/anonymize-client-app-for-demo.md skills/learned-css-that-never-applied.md skills/learned-detection-gaps.md skills/learned-local-db-confusion.md skills/full-harness-audit.md skills/test-result-audit.md skills/finding-disposition.md skills/handoff-to-studio-pro.md skills/preview-over-hub-tunnel.md skills/platform-link.md skills/teamserver-alignment.md skills/learned-constants-and-secrets.md" ;;
+    7)  echo "skills/interview-protocol.md skills/change-intake.md skills/change-intake-detail.md skills/grill-mode.md skills/checkpoints/checkpoint-template.md skills/checkpoints/checkpoint-cutover.md skills/close-the-loop.md skills/share-demo-package.md skills/handoff-to-studio-pro.md skills/platform-link.md skills/teamserver-alignment.md skills/deploy-to-sandbox.md" ;;
     *)  echo "" ;;
 # <!-- ROUTING:END -->
   esac
@@ -2333,7 +2333,7 @@ fi
 # never written once, and a gate cannot see a marker nobody wrote. A change log makes the ask
 # countable; whether its counts are right on real projects is not yet known, and a blocking
 # guard built on an unproven count is how this repo has shipped false FAILs before.
-# change-intake.md names the condition under which this becomes blocking.
+# change-intake-detail.md "The mechanical check" names the condition under which this becomes blocking.
 CR_STATUS="PASS"; CR_NOTE=""; CR_ROWS=""
 if [ "$REG_AMBIGUOUS" = "1" ] || [ -z "$REGISTER" ] || [ ! -f "$REGISTER" ]; then
   CR_STATUS="MANUAL"
diff --git a/bin/lib/change-log.sh b/bin/lib/change-log.sh
index 6aefe476..783dc1d8 100644
--- a/bin/lib/change-log.sh
+++ b/bin/lib/change-log.sh
@@ -17,7 +17,7 @@
 # toolkit pull) — both mandatory chain steps. Rows are written by the lead per change-intake.md.
 #
 # WARN-ONLY. gate-check.sh reports what mxtk_change_log_scan counts as a WARN line and never
-# blocks on it. change-intake.md says when that changes, and why not yet.
+# blocks on it. change-intake-detail.md "The mechanical check" says when that changes, and why not yet.
 #
 # Sourced, never executed. Callers are `set -euo pipefail`, so keep this side-effect free.
 
@@ -33,9 +33,17 @@ Feedback and change requests, one row each, written BEFORE any edit (`skills/cha
 EOF
 }
 
+# CRLF (2026-10-13 review of #306): a PROJECT.md saved on Windows ends every line in \r, so
+# '^## +change log *$' never matched — gate-check reported "no section" and sync-project.sh
+# appended a SECOND empty one; the scan likewise read 0 rows. Both readers drop a trailing \r
+# first, as gate-check.sh does for every table it reads (tr -d '\r'). awk, not tr | grep -q:
+# under the callers' pipefail an early-exiting grep -q can SIGPIPE the producer and turn a hit
+# into a miss.
+#
 # Usage: mxtk_change_log_has_section <PROJECT.md>   — rc 0 when the heading is present
 mxtk_change_log_has_section() {
-  [ -f "$1" ] && grep -qiE '^## +change log *$' "$1"
+  [ -f "$1" ] || return 1
+  awk '{ sub(/\r$/, "") } tolower($0) ~ /^## +change log *$/ { found = 1; exit } END { exit !found }' "$1"
 }
 
 # Usage: mxtk_change_log_scan <PROJECT.md>
@@ -53,6 +61,7 @@ mxtk_change_log_scan() {
     function trim(s) { gsub(/^[ \t]+|[ \t]+$/, "", s); return s }
     function clean(s) { gsub(/[*`]/, "", s); return trim(s) }
     function owe(id, st, why) { rows = rows sprintf("ROW\t%s\t%s\t%s\t%s\n", id, st, why, ask) }
+    { sub(/\r$/, "") }                          # CRLF register: see mxtk_change_log_has_section
     /<!--/ { incomment = 1 }
     incomment { if ($0 ~ /-->/) incomment = 0; next }
     /^## / { insec = (tolower($0) ~ /^## +change log *$/); next }
diff --git a/bin/lib/skill-routing.tsv b/bin/lib/skill-routing.tsv
index 200b80e4..2299fa62 100644
--- a/bin/lib/skill-routing.tsv
+++ b/bin/lib/skill-routing.tsv
@@ -74,6 +74,7 @@ skills-over-scripts	skills/skills-over-scripts.md	Before writing any .js or .sh
 degrade-to-judgement	skills/degrade-to-judgement.md	Any pass whose input is missing, stale or unresolvable — before recording UNMEASURED, N/A or a silent skip: name what was missing, say what you assessed against instead, still deliver a verdict	all	-	baseline	spine
 interview-protocol	skills/interview-protocol.md	Putting a question TO the user — any gate, any stage: ask in chat not in a file, two named options plus your recommendation, one batch per gate then end the turn	lead,ba,architect	P,0,1,2,3,4,5,6,7	baseline	spine
 change-intake	skills/change-intake.md	The user gives feedback or asks for a change mid-project — "change X", "it also needs Y", a remark on a screenshot or the running app, a stakeholder's ask relayed in chat: log a CR row in PROJECT.md ## Change log BEFORE any edit, size it, then edit the highest artifact it alters (BRD, decision, architecture, wireframe), never the MDL first	lead,ba	2,3,4,5,6,7	baseline	spine
+change-intake-detail	skills/change-intake-detail.md	Applying or proving a change request (CR row), a change whose class, size or status is unclear, or a "Change log:" gate line to explain: full class/size/ask tables, the wrong-vs-right example, apply and prove steps, the owners table and what the warn-only gate check counts	lead,ba,architect	2,3,4,5,6,7	ondemand	spine
 grill-mode	skills/grill-mode.md	Deep, adaptive interview on one topic, on demand, when a checkpoint's 2+1 or a single question batch isn't enough	ba,architect	P,0,1,2,3,4,5,6,7	ondemand	spine
 checkpoints	skills/checkpoints/checkpoint-template.md	Any stage transition — the 2+1 format every CAC uses, and the one-register rule (answers land in PROJECT.md, never in a separate state file). The seven CACs themselves are routed per stage in the situational table	lead,ba,architect	0,1,2,3,4,6,7	baseline	spine
 checkpoint-scope	skills/checkpoints/checkpoint-scope.md	CAC-1, closing Stage 0 in EVERY entry mode — scope IN: full scope or a slice, and in what order. Opens with a brainstorm, not options	ba,architect	0	ondemand	spine
diff --git a/skills/agent-roles.md b/skills/agent-roles.md
index 01ba7ba4..c730f06e 100644
--- a/skills/agent-roles.md
+++ b/skills/agent-roles.md
@@ -111,7 +111,7 @@ A single do-everything agent has no natural place to stop before mutating the re
 | **test-agent** | Walk UI test scenarios against the running app, cross-check against the database | No | No — read-only/verification-only |
 | **review-agent** | Owns `module-review.md` — the one pass that closes a module. Stage 3's model-side instruments (`bin/review-module.sh`: conformance + graph sweep) **and stage 4, the LOOK**: every page in the module assessed against wireframe, design system and plain design judgement. Diagnostic only — never fixes what it finds | No — never | Yes, but only its own report under `design/ui-reviews/`, via `Bash` |
 
-**Who records a change, who flushes it.** A mid-build change or decision is *marked* by the lead: the `PROJECT.md` `## Change log` row, the decision row, any `[sync:]` marker. `mdl-agent` and `gate-agent` report it and never write `PROJECT.md`. `ba-agent` flushes BRDs and module briefs, and `architect-agent` flushes wireframes and `architecture/`. The single source is the Owners table in `change-intake.md`.
+**Who records a change, who flushes it.** A mid-build change or decision is *marked* by the lead: the `PROJECT.md` `## Change log` row, the decision row, any `[sync:]` marker. `mdl-agent` and `gate-agent` report it and never write `PROJECT.md`. `ba-agent` flushes BRDs and module briefs, and `architect-agent` flushes wireframes and `architecture/`. The single source is the Owners section in `change-intake.md`.
 
 **The one rule that matters more than any other: only the main session (the one talking to the user) ever runs the command that mutates the real `.mpr`.** All six subagents get `tools: Read, Grep, Glob, Bash` — never `Write`/`Edit`. Agents that need to produce files (`ba-agent`, `architect-agent`, `mdl-agent`) do so via `Bash` (shell redirection/heredoc), not the `Write` tool — that keeps "can this agent silently overwrite something outside its lane" a visible, auditable choice in its tool list rather than an assumption.
 
diff --git a/skills/change-intake-detail.md b/skills/change-intake-detail.md
new file mode 100644
index 00000000..fca1c690
--- /dev/null
+++ b/skills/change-intake-detail.md
@@ -0,0 +1,150 @@
+# Change Intake — Detail: Why, Worked Example, Apply, Prove, Gate Check
+
+**Applies to:** any mxcli project with a `PROJECT.md` (migration, requirements-driven, greenfield, change an existing app), from Stage 2 onward.
+**Purpose:** the on-demand half of `change-intake.md`. That file is the baseline core every lead and ba-agent session reads (trigger, row format, class → entry artifact, size → ask, owners, Done when). Read this one when you apply or prove a change request, when a class, size or status is unclear, or when the `Change log:` gate line needs explaining.
+**Owns:** nothing on its own: `change-intake.md` owns the `## Change log` table in `PROJECT.md` and the classes, sizes and statuses. `conversion-runbook.md` §3b (drift-sync) is the BRD-flush step of this loop; `iterative-build-loop.md`, `close-the-loop.md` and `agent-roles.md` point at `change-intake.md`.
+
+**Trigger:** any message from the user that asks for something to be different from what the artifacts say or what was built, typed, pasted, relayed ("the finance lead wants..."), or said while looking at a screenshot or the running app. Not a trigger: answering a question you asked at a gate (that is `interview-protocol.md`), or a finding from a review pass (that is `finding-disposition.md`).
+
+---
+
+## Why this exists: the change went into MDL, every time
+
+Across 12 mid-build changes on real projects:
+
+| What happened | Count |
+|---|---|
+| Change applied straight in an MDL script | 12 of 12 |
+| BRD, build plan or wireframe updated for it | 0 of 12 |
+| `[sync: … UNSYNCED]` drift markers ever written | 0 |
+| One screenshot-driven fix: rounds before it stuck | 4 (it regressed twice, and every gate was green both times) |
+| One settled entity change sent through ba-agent twice and architect-agent once | ~270k tokens for a handful of document edits |
+
+The drift gate passed on all 12, because it looks for a marker and nobody wrote one. A missing marker looks the same as nothing to sync. The second failure is the opposite one: too much ceremony for a change that was already settled. This skill fixes the first with a row you write *before* the edit. It fixes the second by scaling the ceremony to the size of the change.
+
+### Wrong vs right, same request
+
+The user, looking at the running app: *"The customer form also needs a VAT number."*
+
+**Wrong (what happened 12 times):**
+```
+Sure — adding it now.
+  → mdlsource/0412-customer-vat.mdl   ALTER ENTITY + ALTER PAGE, exec, green
+PROJECT.md: no row.  F003.brd.json: no VAT.  wireframe: no field.  build plan: no step.
+Two sessions later a BRD-driven rebuild of the Customer page drops the field. Every gate is green.
+```
+
+**Right:**
+```
+Logged CR-07: requirement, small → F003.brd.json + Customer_Edit, then MDL.
+  1. PROJECT.md ## Change log: CR-07 row, status agreed (user typed it, small)
+  2. F003.brd.json: VATNumber attribute + its validation, a one-leaf edit by the lead
+  3. Customer_Edit page built → before/after screenshot pair in design/ui-reviews/
+  4. ./bin/exec.sh --cr CR-07 mdlsource/0412-customer-vat.mdl
+  5. Journey customer-edit asserts the field; CR-07 → proven, Closed-by: <commit>, journey customer-edit
+```
+The extra cost is two lines of text and one BRD leaf. The next agent that reads F003 gets VAT.
+
+---
+
+## Step 1 — Capture: one row, before any edit
+
+Append a row to `PROJECT.md` → `## Change log` (the one decision register; no separate file). If the section is missing, `bin/sync-project.sh <project-root>` adds it.
+
+```
+| CR | Date | Source | Verbatim ask | Class | Size | Entry artifact | Status | Closed-by |
+|---|---|---|---|---|---|---|---|---|
+| CR-07 | 2026-10-13 | user | The customer form also needs a VAT number | requirement | S | F003.brd.json | agreed | |
+```
+
+- **CR**: `CR-` plus the next number. Never reuse a number.
+- **Source**: `user`, or the stakeholder's name and role as the user gave it (`stakeholder (finance lead)`).
+- **Verbatim ask**: the words as typed, not your summary. Your restatement goes in the reply (Step 3). Escape a pipe as `\|`.
+- One ask, one row. "Also make the button blue and add a VAT field" is two rows.
+
+## Step 2 — Classify and size
+
+**Class → entry artifact.** The entry artifact is the highest layer the change alters. It is never the MDL.
+
+| Class | What changed | Entry artifact |
+|---|---|---|
+| `requirement` | what the app must do or store | the BRD (`*.brd.json`), plus a decision row if it settles a choice |
+| `decision` | a choice recorded in `PROJECT.md` | a **new** `CONFIRMED` Decisions row whose Notes say `supersedes D-NN`, or `supersedes <stage>: <old decision, first words>` if the register does not number its decisions. The old row's Status becomes `SUPERSEDED by CR-NN`. Its text is never edited |
+| `structure` | module boundary, entity ownership, integration, security model | `architecture/` blueprint or the module brief |
+| `visual` | how a page looks or flows | the wireframe while the page is unbuilt. Once it is built: a before/after screenshot pair in `design/ui-reviews/`, and the wireframe gets a "BUILT, see ui-reviews" stamp instead of upkeep |
+| `impl` | how something is built, with no behaviour change | none: a regression journey only |
+| `bug` | the build does not do what an artifact already says | none: the artifact was right. Add a regression journey that fails before the fix |
+
+**Size.**
+
+| Size | Test (all must hold for S) |
+|---|---|
+| **S** | one artifact per layer, one module, contradicts no `CONFIRMED` decision |
+| **M** | several artifacts in a layer, or 2+ modules. Still contradicts nothing `CONFIRMED` |
+| **L** | a new feature or module, or it contradicts a `CONFIRMED` decision |
+
+## Step 3 — Who is asked, and when (ask, then stop)
+
+Asking follows `interview-protocol.md` §3. On Claude Code use `AskUserQuestion`. On any other agent use numbered markdown, as §3 "Asking on a non-Claude agent" describes. Either way, end the turn and wait.
+
+| Situation | Ask? | Status path | Who edits |
+|---|---|---|---|
+| **S**, typed by the user themselves | **No question.** Reply with a one-line receipt: `Logged CR-07: visual, small → wireframe + page` | straight to `agreed` | the lead, inline. No subagent |
+| **M** | Restate it in your reply (`refined`). Then **one batched question at the next module gate**: "Apply these N changes? CR-08, CR-09, CR-11" | `open` → `refined` → `agreed` at the gate | one `ba-agent` pass at the gate for the whole batch |
+| **L**, or it contradicts a `CONFIRMED` decision | **Ask and stop now**, before anything is built: the proposal, two named options, your recommendation | `refined` → `agreed` / `rejected` | re-enter the pipeline at the owning stage for that slice (Stage 2 for a new BRD, Stage 3 for structure) |
+| Stakeholder or relayed feedback, any size | Not on its own. It stays `open` until the user agrees or rejects. All such rows are surfaced together at the next gate | `open` → `agreed` / `rejected` | as for its size, after agreement |
+| The meaning is ambiguous | **One** "A or B?" question. It is the only mid-stream interruption this skill allows | `open` until answered | — |
+
+A row the user rejects is kept, with Status `rejected — <reason>`. A rejected ask that nobody can find gets asked for again.
+
+## Step 4 — Apply: top down, as far down as needed
+
+1. Edit the entry artifact (Step 2).
+2. Cascade down only to layers that *restate* what changed. Module briefs point to the BRD and do not copy it (`module-brief.md`), so most BRD edits need no brief edit. If the change adds or retires a build step, add or edit that row in the build plan.
+3. Then write the MDL and run it with `./bin/exec.sh --cr CR-NN <script>`. The tag lands in the BUILD-LOG detail cell, and that BUILD-LOG row is the evidence for Step 5.
+4. Set the status to `applied` and fill Closed-by with the commit that holds the artifact edit.
+
+**Owners**, expanded. The source is `change-intake.md` → Owners; this table only spells it out (it replaces the split that `agent-roles.md` lanes contradicted):
+
+| Duty | Owner |
+|---|---|
+| Write and update the CR row and any `PROJECT.md` decision row (the "mark") | the **lead**, the session talking to the user. Helpers report a change to the lead. They never write `PROJECT.md` from the build lanes |
+| Flush a BRD or a module brief (an M batch, or any deferred `[sync: … UNSYNCED]` marker) | **ba-agent** |
+| Flush a wireframe, `design/`, or `architecture/` | **architect-agent** (it owns `design/`) |
+| S edits to any of the above | the **lead**, inline |
+
+When the lead edits a BRD inline in the same session, no `UNSYNCED` marker is needed, because nothing was left unsynced. Use the marker (runbook §3b) only when a flush is deferred.
+
+## Step 5 — Prove
+
+Set the status to `proven` once Closed-by cites the evidence: the commit, plus the journey, test or screenshot pair that shows the change working. For a `bug` or an `impl` row, that is the regression journey. Never weaken or delete an existing journey to make a change pass. That needs its own CR row, which the user agrees to.
+
+---
+
+## The mechanical check (warn-only for now)
+
+`bin/gate-check.sh` prints a `Change log:` line next to `Drift (BRD sync):` and a row on the dashboard:
+
+```
+Change log: WARN — 2 of 6 change request(s) not yet applied (open 1, refined 0, agreed 1); 1 applied/proven with no Closed-by — warn-only (skills/change-intake.md)
+  CR-02  applied  no Closed-by evidence — Orders overview should be a gallery, not a grid
+```
+
+It counts `open`, `refined` and `agreed` rows, plus `applied`/`proven` rows with an empty Closed-by. It also flags any status outside the six words. It **never blocks.** It becomes blocking, for `agreed` rows still unapplied at a module gate, once two projects have run with it and its counts matched a hand count of their change logs. Until then a false FAIL would cost more than the WARN does. What it cannot see is an ask that was never logged: `0 change requests logged` is not proof that nothing changed. That gap is closed by the completion criteria below, not by the script.
+
+When nothing is owed the line is `PASS`, in one of two forms:
+
+```
+Change log: PASS — 0 change requests logged (a change made with no row is invisible here — log every ask first, skills/change-intake.md)
+Change log: PASS — 6 change request(s); every one applied with evidence, proven, or rejected
+```
+
+The first is not evidence that nothing changed (see above). The second is what `change-intake.md` → Done when means by "reads `PASS`".
+
+## Checking each Done-when line
+
+- **Asks → rows:** every change the user asked for in this session has a CR row. N asks give N rows. Count them in the transcript before you claim the module gate.
+- **No MDL before the row:** every `mdlsource/` script written for a CR was executed with `--cr CR-NN`. Its BUILD-LOG row shows `cr: CR-NN`.
+- **Gate line:** before a module gate, `Change log:` reads `PASS` (`N change request(s); every one applied with evidence, proven, or rejected`, or `0 change requests logged`), or every row it still lists under a `WARN` is named in your gate message with a reason.
+- **Evidence:** every `applied`/`proven` row has a Closed-by. Zero rows reported "no Closed-by evidence".
+- **Superseded, not edited:** every `decision` row has a new Decisions row. The old row's Status reads `SUPERSEDED by CR-NN`.
diff --git a/skills/change-intake.md b/skills/change-intake.md
index 2e0cb952..dc39eb30 100644
--- a/skills/change-intake.md
+++ b/skills/change-intake.md
@@ -1,141 +1,39 @@
-# Change Intake — Feedback Typed in the Terminal, Logged Before It Is Built
+# Change Intake — Log Feedback Before You Build It
 
-**Applies to:** any mxcli project with a `PROJECT.md` (migration, requirements-driven, greenfield, change an existing app), from Stage 2 onward.
-**Purpose:** What to do the moment the user, or a stakeholder they relay, says "change X", "this also needs Y", "that page looks wrong": log it, size it, agree it at the right moment, edit the highest artifact it alters, and prove it. The MDL comes last, never first.
-**Owns:** the `## Change log` table in `PROJECT.md` and the classes, sizes and statuses below. `conversion-runbook.md` §3b (drift-sync) is the BRD-flush step of this loop and points here; `iterative-build-loop.md`, `close-the-loop.md` and `agent-roles.md` point here too.
+**Applies to:** any project with `PROJECT.md`, Stage 2 onward.
+**Purpose:** log a mid-project change before any edit, then edit the highest artifact it alters; MDL last. Why, example, apply/prove, gate check: `change-intake-detail.md`.
 
-**Trigger:** any message from the user that asks for something to be different from what the artifacts say or what was built, typed, pasted, relayed ("the finance lead wants..."), or said while looking at a screenshot or the running app. Not a trigger: answering a question you asked at a gate (that is `interview-protocol.md`), or a finding from a review pass (that is `finding-disposition.md`).
+**Trigger:** the user, or a relayed stakeholder, asks for anything to differ from artifacts or build. Not a gate answer or review finding.
 
----
+Wrong: "Sure — adding it now" → MDL, no row (12 of 12 field changes).
 
-## Why this exists: the change went into MDL, every time
+## Row first, in `PROJECT.md` `## Change log`
 
-Across 12 mid-build changes on real projects:
+Columns: CR-NN (never reused), Date, Source, Verbatim ask (`\|` for a pipe), Class, Size, Entry artifact, Status, Closed-by. One ask, one row. Status: open → refined → agreed → applied → proven, or `rejected — reason` (kept).
 
-| What happened | Count |
-|---|---|
-| Change applied straight in an MDL script | 12 of 12 |
-| BRD, build plan or wireframe updated for it | 0 of 12 |
-| `[sync: … UNSYNCED]` drift markers ever written | 0 |
-| One screenshot-driven fix: rounds before it stuck | 4 (it regressed twice, and every gate was green both times) |
-| One settled entity change sent through ba-agent twice and architect-agent once | ~270k tokens for a handful of document edits |
+## Class → entry artifact (never the MDL)
 
-The drift gate passed on all 12, because it looks for a marker and nobody wrote one. A missing marker looks the same as nothing to sync. The second failure is the opposite one: too much ceremony for a change that was already settled. This skill fixes the first with a row you write *before* the edit. It fixes the second by scaling the ceremony to the size of the change.
+- **requirement** → BRD.
+- **decision** → new `CONFIRMED` row `supersedes D-NN`; old row's Status `SUPERSEDED by CR-NN`, text untouched.
+- **structure** → `architecture/` or module brief.
+- **visual** → wireframe while unbuilt; once built, before/after pair in `design/ui-reviews/`.
+- **impl**, **bug** → regression journey (a bug's fails first).
 
-### Wrong vs right, same request
+## Size → ask (ask, then stop)
 
-The user, looking at the running app: *"The customer form also needs a VAT number."*
+- **S** (one artifact per layer, one module, contradicts nothing `CONFIRMED`), user-typed: no question; receipt `Logged CR-07: visual, S → wireframe`, `agreed`, lead edits.
+- **M** (several artifacts or modules): restate (`refined`); one batched question at the module gate, one ba-agent pass.
+- **L** (new feature or module, or contradicts `CONFIRMED`): ask now (two options, a recommendation); re-enter the owning stage.
+- **Relayed**: `open` until the next gate asks. **Ambiguous**: one "A or B?".
 
-**Wrong (what happened 12 times):**
-```
-Sure — adding it now.
-  → mdlsource/0412-customer-vat.mdl   ALTER ENTITY + ALTER PAGE, exec, green
-PROJECT.md: no row.  F003.brd.json: no VAT.  wireframe: no field.  build plan: no step.
-Two sessions later a BRD-driven rebuild of the Customer page drops the field. Every gate is green.
-```
+## Owners (single source)
 
-**Right:**
-```
-Logged CR-07: requirement, small → F003.brd.json + Customer_Edit, then MDL.
-  1. PROJECT.md ## Change log: CR-07 row, status agreed (user typed it, small)
-  2. F003.brd.json: VATNumber attribute + its validation, a one-leaf edit by the lead
-  3. Customer_Edit page built → before/after screenshot pair in design/ui-reviews/
-  4. ./bin/exec.sh --cr CR-07 mdlsource/0412-customer-vat.mdl
-  5. Journey customer-edit asserts the field; CR-07 → proven, Closed-by: <commit>, journey customer-edit
-```
-The extra cost is two lines of text and one BRD leaf. The next agent that reads F003 gets VAT.
+Lead: all CR/decision rows, S edits; helpers never write `PROJECT.md`. ba-agent: BRDs, briefs. architect-agent: wireframes, `design/`, `architecture/`. When the lead edits inline, no `UNSYNCED` marker is needed; use one only for a deferred flush.
 
----
+## Done when
 
-## Step 1 — Capture: one row, before any edit
-
-Append a row to `PROJECT.md` → `## Change log` (the one decision register; no separate file). If the section is missing, `bin/sync-project.sh <project-root>` adds it.
-
-```
-| CR | Date | Source | Verbatim ask | Class | Size | Entry artifact | Status | Closed-by |
-|---|---|---|---|---|---|---|---|---|
-| CR-07 | 2026-10-13 | user | The customer form also needs a VAT number | requirement | S | F003.brd.json | agreed | |
-```
-
-- **CR**: `CR-` plus the next number. Never reuse a number.
-- **Source**: `user`, or the stakeholder's name and role as the user gave it (`stakeholder (finance lead)`).
-- **Verbatim ask**: the words as typed, not your summary. Your restatement goes in the reply (Step 3). Escape a pipe as `\|`.
-- One ask, one row. "Also make the button blue and add a VAT field" is two rows.
-
-## Step 2 — Classify and size
-
-**Class → entry artifact.** The entry artifact is the highest layer the change alters. It is never the MDL.
-
-| Class | What changed | Entry artifact |
-|---|---|---|
-| `requirement` | what the app must do or store | the BRD (`*.brd.json`), plus a decision row if it settles a choice |
-| `decision` | a choice recorded in `PROJECT.md` | a **new** `CONFIRMED` Decisions row whose Notes say `supersedes D-NN`, or `supersedes <stage>: <old decision, first words>` if the register does not number its decisions. The old row's Status becomes `SUPERSEDED by CR-NN`. Its text is never edited |
-| `structure` | module boundary, entity ownership, integration, security model | `architecture/` blueprint or the module brief |
-| `visual` | how a page looks or flows | the wireframe while the page is unbuilt. Once it is built: a before/after screenshot pair in `design/ui-reviews/`, and the wireframe gets a "BUILT, see ui-reviews" stamp instead of upkeep |
-| `impl` | how something is built, with no behaviour change | none: a regression journey only |
-| `bug` | the build does not do what an artifact already says | none: the artifact was right. Add a regression journey that fails before the fix |
-
-**Size.**
-
-| Size | Test (all must hold for S) |
-|---|---|
-| **S** | one artifact per layer, one module, contradicts no `CONFIRMED` decision |
-| **M** | several artifacts in a layer, or 2+ modules. Still contradicts nothing `CONFIRMED` |
-| **L** | a new feature or module, or it contradicts a `CONFIRMED` decision |
-
-## Step 3 — Who is asked, and when (ask, then stop)
-
-Asking follows `interview-protocol.md` §3. On Claude Code use `AskUserQuestion`. On any other agent use numbered markdown, as §3 "Asking on a non-Claude agent" describes. Either way, end the turn and wait.
-
-| Situation | Ask? | Status path | Who edits |
-|---|---|---|---|
-| **S**, typed by the user themselves | **No question.** Reply with a one-line receipt: `Logged CR-07: visual, small → wireframe + page` | straight to `agreed` | the lead, inline. No subagent |
-| **M** | Restate it in your reply (`refined`). Then **one batched question at the next module gate**: "Apply these N changes? CR-08, CR-09, CR-11" | `open` → `refined` → `agreed` at the gate | one `ba-agent` pass at the gate for the whole batch |
-| **L**, or it contradicts a `CONFIRMED` decision | **Ask and stop now**, before anything is built: the proposal, two named options, your recommendation | `refined` → `agreed` / `rejected` | re-enter the pipeline at the owning stage for that slice (Stage 2 for a new BRD, Stage 3 for structure) |
-| Stakeholder or relayed feedback, any size | Not on its own. It stays `open` until the user agrees or rejects. All such rows are surfaced together at the next gate | `open` → `agreed` / `rejected` | as for its size, after agreement |
-| The meaning is ambiguous | **One** "A or B?" question. It is the only mid-stream interruption this skill allows | `open` until answered | — |
-
-A row the user rejects is kept, with Status `rejected — <reason>`. A rejected ask that nobody can find gets asked for again.
-
-## Step 4 — Apply: top down, as far down as needed
-
-1. Edit the entry artifact (Step 2).
-2. Cascade down only to layers that *restate* what changed. Module briefs point to the BRD and do not copy it (`module-brief.md`), so most BRD edits need no brief edit. If the change adds or retires a build step, add or edit that row in the build plan.
-3. Then write the MDL and run it with `./bin/exec.sh --cr CR-NN <script>`. The tag lands in the BUILD-LOG detail cell, and that BUILD-LOG row is the evidence for Step 5.
-4. Set the status to `applied` and fill Closed-by with the commit that holds the artifact edit.
-
-**Owners** (this replaces the split that `agent-roles.md` lanes contradicted):
-
-| Duty | Owner |
-|---|---|
-| Write and update the CR row and any `PROJECT.md` decision row (the "mark") | the **lead**, the session talking to the user. Helpers report a change to the lead. They never write `PROJECT.md` from the build lanes |
-| Flush a BRD or a module brief (an M batch, or any deferred `[sync: … UNSYNCED]` marker) | **ba-agent** |
-| Flush a wireframe, `design/`, or `architecture/` | **architect-agent** (it owns `design/`) |
-| S edits to any of the above | the **lead**, inline |
-
-When the lead edits a BRD inline in the same session, no `UNSYNCED` marker is needed, because nothing was left unsynced. Use the marker (runbook §3b) only when a flush is deferred.
-
-## Step 5 — Prove
-
-Set the status to `proven` once Closed-by cites the evidence: the commit, plus the journey, test or screenshot pair that shows the change working. For a `bug` or an `impl` row, that is the regression journey. Never weaken or delete an existing journey to make a change pass. That needs its own CR row, which the user agrees to.
-
----
-
-## The mechanical check (warn-only for now)
-
-`bin/gate-check.sh` prints a `Change log:` line next to `Drift (BRD sync):` and a row on the dashboard:
-
-```
-Change log: WARN — 2 of 6 change request(s) not yet applied (open 1, refined 0, agreed 1); 1 applied/proven with no Closed-by — warn-only (skills/change-intake.md)
-  CR-02  applied  no Closed-by evidence — Orders overview should be a gallery, not a grid
-```
-
-It counts `open`, `refined` and `agreed` rows, plus `applied`/`proven` rows with an empty Closed-by. It also flags any status outside the six words. It **never blocks.** It becomes blocking, for `agreed` rows still unapplied at a module gate, once two projects have run with it and its counts matched a hand count of their change logs. Until then a false FAIL would cost more than the WARN does. What it cannot see is an ask that was never logged: `0 change requests logged` is not proof that nothing changed. That gap is closed by the completion criteria below, not by the script.
-
-## Done when (denominators)
-
-- **Asks → rows:** every change the user asked for in this session has a CR row. N asks give N rows. Count them in the transcript before you claim the module gate.
-- **No MDL before the row:** every `mdlsource/` script written for a CR was executed with `--cr CR-NN`. Its BUILD-LOG row shows `cr: CR-NN`.
-- **Gate line:** before a module gate, `Change log:` reads `0 of M … not yet applied`, or every remaining row is named in your gate message with a reason.
-- **Evidence:** every `applied`/`proven` row has a Closed-by. Zero rows reported "no Closed-by evidence".
-- **Superseded, not edited:** every `decision` row has a new Decisions row. The old row's Status reads `SUPERSEDED by CR-NN`.
+- N asks → N rows.
+- Every CR script ran via `./bin/exec.sh --cr CR-NN`.
+- At a module gate, `Change log:` reads `PASS`, or each listed row is named with a reason.
+- Zero "no Closed-by evidence" rows.
+- Decisions superseded, not edited.
diff --git a/skills/conversion-runbook.md b/skills/conversion-runbook.md
index cc80daa8..32b07cdc 100644
--- a/skills/conversion-runbook.md
+++ b/skills/conversion-runbook.md
@@ -805,7 +805,7 @@ stale assertion. This is a real, observed failure mode, not a hypothetical — s
 `iterative-build-loop.md` → "Requirements Drift-Sync Rule" (where the detection mechanics — which
 BRD fields count as "asserted" — live).
 
-**Ownership is split, on purpose** (the table in `change-intake.md` → Owners is the single source):
+**Ownership is split, on purpose** (`change-intake.md` → Owners is the single source; `change-intake-detail.md` expands it):
 - **The lead** (the session talking to the user) **marks**: it writes the decision row and its
   marker. A helper that hits a BRD-touching decision reports it to the lead, because `mdl-agent`
   writes only `.mdl` and `gate-agent` is read-only (`agent-roles.md`).
diff --git a/tests/wave2/test-change-log.sh b/tests/wave2/test-change-log.sh
index 0755268e..0c5c7ab6 100755
--- a/tests/wave2/test-change-log.sh
+++ b/tests/wave2/test-change-log.sh
@@ -22,11 +22,16 @@
 #       --dry-run writes nothing
 #   T8  exec.sh --cr: a malformed id is refused (exit 2) before anything needs a project; a
 #       well-formed one passes through to --change-class; the BUILD-LOG detail carries "cr:"
+#   T9  CRLF PROJECT.md (saved on Windows): the section is found, the scan and the gate line
+#       match the LF copy exactly, and sync does not append a second section
 
 export MXTK_LEAKGUARD_DENYFILE="${TMPDIR:-/tmp}/mxtk-fixture-denylist.$$"
 export MXTK_NO_GUIDE=1
 export MXTK_SYNC_SKIP_CLONE_CHECK=1
-set -uo pipefail
+# No pipefail, on purpose: every pipeline here is `printf "$OUT" | grep -q …`, and under pipefail a
+# grep -q that matches early can SIGPIPE printf on a long $OUT, turning a hit into a FAIL. Seen as
+# an intermittent "no sync message" in T7 (2026-10-13). No assertion reads an upstream exit code.
+set -u
 
 GATE="${1:?usage: test-change-log.sh /path/to/gate-check.sh}"
 case "$GATE" in /*) ;; *) GATE="$PWD/$GATE" ;; esac
@@ -188,6 +193,37 @@ OUT="$(cd "$WORK/empty" && bash "$EXEC" --cr CR-07 --change-class "$MDL" 2>&1)";
   || bad "--cr then --change-class: rc=$RC" "$OUT"
 grep -q 'cr: \$CR_ID' "$EXEC" && ok "BUILD-LOG detail carries 'cr: <id>'" || bad "BUILD-LOG detail has no cr tag"
 
+echo "== T9: CRLF register =="
+# Same rows as T2 on a fresh real scaffold, then every line ended in CRLF — what an editor on
+# Windows writes. Before the fix the heading regex never matched: has_section failed, the scan
+# counted 0 rows, gate-check said "no section", and sync-project.sh appended a second one.
+P9="$(mkproj t9)"; R9="$P9/PROJECT.md"
+add_rows "$R9" <<'EOF2'
+| CR-01 | 2026-10-13 | user | Customer also needs a VAT number | requirement | S | F003.brd.json | proven | abc1234, journey customer-edit |
+| CR-02 | 2026-10-13 | user | Orders overview should be a gallery, not a grid | visual | S | design/ui-reviews/orders | applied | |
+| CR-03 | 2026-10-13 | stakeholder (finance lead) | Approvals over 10k need a second approver | decision | L | PROJECT.md new row | open | |
+| CR-04 | 2026-10-13 | user | Search \| Reset buttons do nothing | bug | S | none | **agreed** | |
+EOF2
+LF_SCAN="$(mxtk_change_log_scan "$R9")"
+LF_LINE="$("$GATE" --no-html "$P9" 2>&1 | grep '^Change log:')"
+case "$LF_LINE" in "Change log: WARN — 2 of 4 change request(s) not yet applied (open 1, refined 0, agreed 1); 1 applied/proven with no Closed-by"*) ok "LF baseline: WARN, 2 of 4 pending + 1 without evidence" ;;
+  *) bad "LF baseline line" "$LF_LINE" ;; esac
+awk '{ printf "%s\r\n", $0 }' "$R9" > "$R9.crlf" && mv "$R9.crlf" "$R9"
+grep -q $'\r$' "$R9" && ok "precondition: PROJECT.md is CRLF" || bad "CRLF conversion failed"
+mxtk_change_log_has_section "$R9" && ok "has_section finds the heading in a CRLF file" || bad "has_section misses the CRLF heading"
+CR_SCAN="$(mxtk_change_log_scan "$R9")"
+[ "$CR_SCAN" = "$LF_SCAN" ] && ok "scan of the CRLF file = scan of the LF file" \
+  || bad "CRLF scan differs" "$(printf 'LF:\n%s\nCRLF:\n%s' "$LF_SCAN" "$CR_SCAN" | cat -A)"
+printf '%s\n' "$CR_SCAN" | grep -q $'\r' && bad "a \\r leaked into the scan output" || ok "no \\r in the scan output"
+CR_LINE="$("$GATE" --no-html "$P9" 2>&1 | grep '^Change log:')"
+[ "$CR_LINE" = "$LF_LINE" ] && ok "gate line on CRLF = gate line on LF ($LF_LINE)" \
+  || bad "gate line differs on CRLF" "$(printf 'LF:   %s\nCRLF: %s' "$LF_LINE" "$CR_LINE")"
+OUT="$("$SYNC" "$P9" 2>&1)"
+[ "$(count_heading "$R9")" = "1" ] && ok "sync on CRLF: still exactly one '## Change log'" \
+  || bad "sync on CRLF: $(count_heading "$R9") headings"
+printf '%s\n' "$OUT" | grep -q "appended the '## Change log' section" && bad "sync claims it appended a section to a CRLF file that has one" \
+  || ok "sync does not append to a CRLF file"
+
 echo ""
 echo "test-change-log: $PASS passed, $FAIL failed"
 [ "$FAIL" -eq 0 ]