fix(fr): align /arckit:fr-irn with the IRN v1.2 grid

[thomas-jardinet]

Why

/arckit:fr-irn still cited IRN v0.4 and Questionnaire_IRN_v0.4.xlsx. The aDRI repository now publishes Référentiel_IRN_v1.2.xlsx (uploaded 2026-09-22; v1.0 landed 2026-06-09), and the grid itself changed shape:

	v0.4 (what ArcKit scaffolded)	v1.2 workbook
Unit of assessment	8 pillars × 5 organisational layers	criteria RES-N.n, filled once per digital asset
Criterion scope	—	Organisation or Actif numérique
Scale	R / NR	maturity level, or non-resilient / not applicable, with comments and evidence
0–100 weighted score	promised per pillar and globally	not computed by the workbook

The five layers do not appear anywhere in the v1.2 grid, so the old scaffold produced a matrix an assessor could not fill from the official file.

What changed

• Command (plugins/arckit-fr/commands/fr-irn.md): points at the v1.2 workbook, scopes by digital asset, refers to criteria by ID only.
• Template (plugins/arckit-fr/templates/ and .arckit/templates/): asset inventory replaces the layer table; each pillar has a per-criterion grid (ID, scope, asset, maturity level, evidence); the summary matrix is pillars × Organisation and each asset; the executive summary no longer promises a 0–100 score.
• Guide (docs/guides/fr-irn.md, synced to the plugin copy): "The 5 Organisational Layers" becomes "How the Grid Is Assessed (IRN v1.2)".
• Quality checklist: the three IRN checks that named layers and R/NR are rewritten; resynced to every plugin, which accounts for most of the 41 files.
• RES-7 and RES-8 take their v1.2 names (Sécurité & Résilience; Résilience Environnementale et Énergétique).
• CHANGELOG entry under [Unreleased].

Deliberately not hard-coded

• Maturity scale labels — the v1.2 workbook carries two different dropdown lists depending on the row, so the template defers to the official grid.
• Criterion count — 34 IDs today, five of them under RES-7 marked pending validation.
• Criterion text — still not reproduced; the licence is unchanged (CC BY-NC-ND 4.0).

Note that the upstream README still describes v0.4, R/NR and the five layers; the workbook is what this PR follows.

Checks

• sync-shared-assets.py --check, check-guide-parity.py: OK
• pytest on test_claude_overlay_namespacing, test_release_process, test_doc_control_resolution, test_quality_checklist_ladders: 90 passed (full suite not run locally)
• markdownlint-cli2 on the edited files: 0 issues
• converter.py: runs clean

🤖 Generated with Claude Code

[tractorjuice]

Code Review

Verdict: Approved. I checked this against Référentiel_IRN_v1.2.xlsx on the aDRI repository, and the PR describes the grid accurately. I've resolved the one conflict on your branch and will merge once CI is green.

BLOCKERS

The content has none. The only issue was a CHANGELOG.md conflict with the 6.17.4 and 6.17.5 sections added since the branch was cut. I pushed a merge commit to your branch that keeps your ### Changed entry under [Unreleased], above ## [6.17.5], and changes nothing else.

Checked against the v1.2 workbook

• Assessment per digital asset. Grille V1 has one column group per asset ("Actif 1", "Actif 2") under seven asset fields (rows 5–11). ✓
• Criterion scope. The Référentiel v1 dropdown is Actif numérique, Organisation. ✓
• Two maturity scales. Most rows use Intentions, Moyens, Résultats, Non résilient, Non concerné, and RES-7.3 to 7.7 use Documenté, Déployé, Contrôlé, Non Concerné, Non Résilient. Deferring to the official grid instead of hard-coding either is the right call. ✓
• No 0–100 score. The workbook contains no formulas, so it computes no score. Dropping that promise from the executive summary is correct. ✓
• Criterion count. There are 34 distinct IDs, and RES-7.3 to RES-7.7 are marked "en attente de validation". ✓
• Pillar names. RES-7 is "Sécurité & Résilience" and RES-8 is "Résilience Environnementale et énergétique". ✓
• Asset importance. The dropdown is Standard, Essentiel, Vital. ✓
• English version. The English workbook is DRI_Referential_v_1.1.xlsx. ✓

IMPORTANT

None.

MINOR (optional, fine as a follow-up)

1. RES-7.3 appears twice upstream. Grid row 40 (optional, pending validation) and row 45 ("Contrôle régional des SOC et journaux", asset scope) carry the same ID. The ArcKit document refers to criteria by ID only, so an assessor copying from the grid gets two RES-7.3 rows they can't tell apart. A short note in the RES-7 section would handle it, for example: "the v1.2 grid uses RES-7.3 twice; label the optional row as optional". It may also be worth raising with the aDRI.
2. There is a third scope value. In Grille V1, the five pending RES-7 rows have Facultatif in "Portée du critère". The command already covers this ("some may be marked optional or pending validation"). The template placeholder [Organisation / Actif numérique] and the guide's "Criterion scope" row list only two values, though, so they could gain / Facultatif.
3. The asset description is missing. The grid has seven asset fields and the inventory table has six. Adding a Description column for "Description de l'actif" (row 7) would let the assessor copy the grid one field to one column.

POSITIVE

• You followed the workbook rather than the upstream README, which still describes v0.4, and the PR description says so.
• The licence handling still holds: criteria appear by ID only, with no titles, descriptions or recommendations.
• The shared quality checklist was resynced correctly, and all three sync layers pass.
• No tracked file still mentions v0.4, R/NR or the five layers, apart from the guide's deliberate note on what changed.

Checks run on this PR merged with current main

• sync-shared-assets.py --check, check-guide-parity.py, sync-claude-plugin-layout.py --check, check-doc-type-registry.py: OK
• converter.py: OK
• pytest tests/: 2283 passed, 228 skipped
• markdownlint-cli2 on the six edited source files: 0 issues

Next

Merging now. The three minor items can follow in a separate PR if you want them.

Thanks for keeping this overlay current, Thomas.