feat(mdl-rules): load MDL rules per statement type instead of preloading (lazy-load PR 2a)

.github/workflows/checks.yml

@@ -33,6 +33,9 @@ jobs:
       - name: Routing tables in sync with skill-routing.tsv (render-routing.sh --check)
         run: bin/render-routing.sh --check
 
+      - name: MDL rules by kind cover every rule and match skills/generated (mdl-rules.sh --check)
+        run: bin/mdl-rules.sh --check
+
       - name: No macOS-only assumptions reintroduced (check-portability.sh)
         run: bin/check-portability.sh
 

CHANGELOG.md

@@ -19,6 +19,7 @@ Sections dated before 2026-09-19 predate the cycle and stay as they are. Named m
 
 ## Unreleased
 
+- new(mdl-rules): **`bin/mdl-rules.sh <kind...> | --for <script.mdl>` prints the MDL rules one script needs, by the kinds of statement it writes** — lazy-load PR 2a. `bin/lib/mdl-rule-kinds.tsv` maps page / microflow / access / workflow / entity / any to sections located by file and heading text (never line numbers): the STOP rows by their `Kind` column, the kind-specific gotchas, `learned-microflow-patterns.md`, `microflow-preflight.md`, `ui-preflight-pages.md`, `design-spacing.md` §1–3, the detection-gaps register rows by construct, `security-is-not-a-later-script.md` and `retesting-learned-rules.md`. Output is the core (Step 0, the `any` STOP rows, gotchas, final self-check: 3,347 words) plus the kind sections, closed by "Rules for kinds K: STOP rows n/n listed" (page 10,223 words, microflow 10,186, access 4,730, entity 5,183, workflow 3,732), against the 29,388 words the mdl helper preloads at Stage 5 today. `--for` reuses exec.sh's statement classifier (`exec.sh --subjects`; Java actions and mappings count as microflow, demo users as access) and exits 1 if it fails; a heading that is gone fails loudly, never prints empty. `skills/generated/rules-<kind>.md` are the same output for harnesses that read files (Copilot, Cursor, Windsurf); `--check` (CI) fails on drift, on any STOP row, gotcha, register row or `##` section that reaches no kind (a section left out on purpose needs an `exclude` row with its reason), and on any register ERE that matches no row by itself, and on an unclosed code fence in a file it reads. No routing change yet: PR 2b demotes the preloads. Behaviour suite: `build-packs.sh --rules "<kinds>"` expands the script in a pack. Fixture `tests/wave2/test-mdl-rules.sh` — lazy-load plan, option C (maintainer)
 - process(preflight): **`learned-mdl-preflight.md` moves its STOP-row narrative out of the baseline and gains a `Kind` column (page / microflow / access / workflow / any)** — lazy-load PR 1. The 15 retired rows collapse to one line listing their numbers, the version notes move, and each live row keeps its If/Then text plus one failure clause; the full rows, retired rows and version notes are in `process/baseline-incident-history.md` § PF1, verbatim (`check-moved-text.sh`: 43 of 43 removed lines found). The workflow `DECISION` STOP and the #879 read-first pointer stay in the skill. `Kind` sits at the end of each row so `mdl-scan`'s `| <n> 🔎` markers still match. Baseline 73,946 → 72,891 words (−1,055). — maintainer (Option C lazy-loading plan)
 - fix(progress-report): **`progress-report.py` no longer crashes with `TypeError: '<' not supported between instances of 'dict' and 'dict'` when two exec'd scripts share a leading number** (#513). `script_phases()` sorted `(number, phase dict)` pairs, so `10-data.mdl` and `10b-data-flows.mdl` (or one script logged under two paths), both matched to a plan phase, compared the dicts and every `exec.sh` run ended with `WARN: progress-report.html not updated`. It now sorts on (number, script name) only; an unmatched numbered script still takes the phase of the nearest lower-numbered matched one, ties broken by name. Fixture: `test-progress-report.sh` T17 (fails 3/3 on master). — jopterhorst (#513)
 - new(canvas): **`project-bin/mf-canvas-facts.py` prints how a microflow or nanoflow lies on the Studio Pro canvas: labels on top of each other, lines through objects, note lines leaving the side that faces away, flows that run right-to-left or curl, captions wider than their shape, and decision captions that ask the opposite of their true branch (SENSE).** It prints facts, one line each, and no verdicts; read-only (it reads units through `mxpatch.py`), exit 2 when a flow could not be read. `verify-module.sh` runs it per module as an info rung (1c, log `14-canvas-facts.log`; a missing script, Python or `.mpr`, an exit 2 or a timeout is an INFO row reading NOT MEASURED, never a gate). Fixture `tests/wave2/test-mf-canvas-facts.sh` runs on a model mxcli v0.27.0 wrote (units and `.mpr` captured); `test-verify-module-446.sh` gains V8 for the rung. Calibration is from Studio Pro 10.24 screenshots, not verified on Studio Pro 11. Field evidence, measured 2026-10-21 on one Mendix 10.24 app: 1,588 flows read in one pass; on the 36 flows an mxcli-built branch added or changed in one module, OVERLAP 19, LINE 27, NOTE-SIDE 17, LEFTWARD 3, TURN 4, SENSE 9 before repair — saingge

bin/lib/mdl-rule-kinds.tsv

@@ -0,0 +1,74 @@
+# mdl-rule-kinds.tsv — which MDL rules a script needs, by the kind of statement it writes.
+# Read by bin/mdl-rules.sh (and nothing else); rendered to skills/generated/rules-<kind>.md.
+#
+# WHY (lazy-load PR 2, 2026-10-23). The mdl helper preloaded every kind-specific skill at Stage 5
+# (29,388 words), though a page script needs none of the microflow rules and a microflow script
+# none of the page rules. This table says which sections each statement kind needs, so the helper
+# loads them per script instead. The skills keep every word; only who preloads them changes.
+#
+# Sections are located by FILE plus HEADING TEXT, never by line number (the bin/lib/runbook-span.sh
+# precedent): a heading matches when its text, after the leading #s, starts with the heading cell.
+# A heading that no longer exists makes bin/mdl-rules.sh fail loudly — it never prints empty.
+#
+# Format: TAB-separated, four columns. Blank lines and #-comments ignored.
+#   1 kind      page | microflow | access | workflow | entity | any   (any = every script)
+#   2 file      repo-relative path
+#   3 selector  file     the whole file
+#               top      everything above the file's first "## " heading
+#               section  the heading and everything under it, down to the next heading of its level
+#               stop     the STOP table under that heading, rows filtered by their Kind column
+#                        (the kinds asked for, plus `any`; the retired-numbers row always)
+#               bullet   one bullet under that heading: arg is "<heading>::<bullet lead text>"
+#               rows     the table rows under that heading whose FIRST cell matches an ERE
+#                        (case-insensitive): arg is "<heading>::<ERE>"
+#               exclude  kind "-": a `##` section deliberately reaching no kind (it is not a rule
+#                        for writing MDL); arg is "<heading>::<one-line reason>"
+#   4 arg       heading text, or heading::lead / heading::ERE / heading::reason as above; "-" for
+#               file and top
+#
+# Denominators (bin/mdl-rules.sh --check): every STOP row carries a known Kind, every bullet under
+# a `bullet` heading is selected by some row here, every table row under a `rows` heading is matched
+# by some ERE here and every ERE matches some row on its own, every `##` section of a file read in
+# parts (no `file` row) is selected or excluded below, and every generated file equals a fresh
+# render. A new gotcha, register row or section that no kind selects fails --check, so it cannot
+# silently stop reaching the mdl helper. Headings inside code fences are not headings.
+#
+# kind	file	selector	arg
+any	skills/learned-mdl-preflight.md	top	-
+any	skills/learned-mdl-preflight.md	section	Step 0 — choose the write mode
+any	skills/learned-mdl-preflight.md	stop	The checklist — run before writing a single line of MDL
+any	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Quote / unquote retry rule:**
+any	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Grants are co-located
+any	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**`mxcli check` does not catch corruption.**
+any	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**mxcli v0.25 — keep the skills' spellings
+any	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**`exec` is not `check`.**
+page	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Cross-module association paths
+microflow	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Cross-module association paths
+page	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Changing an existing DataGrid 2 column
+page	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**No `elseif` in page/widget expressions**
+entity	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Enumeration values named with reserved words**
+access	skills/learned-mdl-preflight.md	bullet	MDL gotchas::**Demo users:**
+any	skills/learned-mdl-preflight.md	section	Final self-check
+page	skills/ui-preflight-pages.md	file	-
+page	skills/design-spacing.md	top	-
+page	skills/design-spacing.md	section	1. The scale
+page	skills/design-spacing.md	section	2. The rhythm hierarchy
+page	skills/design-spacing.md	section	3. The page scaffold
+-	skills/design-spacing.md	exclude	4. Review-side::judging a screenshot at module review (module-review.md 4d), not writing MDL
+-	skills/design-spacing.md	exclude	Related::cross-references to other skills, no rule of its own
+microflow	skills/learned-microflow-patterns.md	file	-
+microflow	skills/microflow-preflight.md	file	-
+access	skills/learned-microflow-patterns.md	section	Current User — Two Separate Mechanisms
+entity	skills/learned-microflow-patterns.md	section	Association Direction — Reading SHOW ASSOCIATIONS Correctly
+access	skills/security-is-not-a-later-script.md	file	-
+entity	skills/security-is-not-a-later-script.md	file	-
+any	skills/learned-detection-gaps.md	top	-
+any	skills/learned-detection-gaps.md	section	The ladder
+any	skills/learned-detection-gaps.md	rows	The register::non-idempotent|mxcli < v0\.26
+page	skills/learned-detection-gaps.md	rows	The register::page|widget|datagrid|dg2|dropdownfilter|navigation|dynamictext|contentparams|elseif|bindings|combo box|snippetcall
+microflow	skills/learned-detection-gaps.md	rows	The register::java action|getvalue|commit|microflow|on error|create entity|xpath|currentuser|import mapping|get workflows|loop
+entity	skills/learned-detection-gaps.md	rows	The register::entity|attribute|enumeration|reserved|calculated by|autocreated
+workflow	skills/learned-detection-gaps.md	rows	The register::workflow|parallel split
+access	skills/learned-detection-gaps.md	rows	The register::access|currentuser|role
+any	skills/learned-detection-gaps.md	section	Operating rules that fall out of this table
+any	skills/retesting-learned-rules.md	file	-

bin/lib/routing-exempt.tsv

@@ -20,3 +20,4 @@ skills/checkpoints/	reached via the checkpoint mechanism (conversion-runbook sta
 skills/journey-map.md	tombstone — falsification trial failed 2026-08-22, row removed then; file kept so old links resolve
 skills/ui-review-loop.md	merge tombstone — content lives in skills/module-review.md (routed)
 skills/module-completion-loop.md	merge tombstone — content lives in skills/module-review.md (routed)
+skills/generated/	rendered output of bin/mdl-rules.sh (rules-<kind>.md, drift-checked by bin/mdl-rules.sh --check) — the plain-read twin of that script's output, reached through bin/mdl-rules.sh's own routing row, never routed one by one