docs: file post-merge residual-sweep queue (conventions stragglers + assistant surfaces)

draft/docs/workspaces/assistants_regime_extension.md

@@ -29,8 +29,33 @@ workflow.
   inherit the multi_galaxy + cluster (light) workflows. Record the
   requirement here; do not create the repo as part of this task.
 
+## Specific stale surfaces (post-merge Opus sweep, 2026-07-25)
+
+The autolens_assistant is systematically one rung behind the shipped split;
+highest-value targets when this task starts:
+
+- `wiki/core/concepts/group_and_cluster_lensing.md` — the canonical
+  above-galaxy page presents a two-rung (group/cluster) set: title, the
+  "galaxy to group to cluster" ladder sentence, and the Einstein-radius
+  enumeration all need the multi_galaxy rung.
+- `wiki/core/concepts/mass_profiles.md` model-picking table — the
+  "Group / cluster lens | multi-galaxy Isothermal ± scaling relations" row
+  uses "multi-galaxy" to MEAN group/cluster; now doubly wrong (the
+  multi_galaxy rung is free per-deflector models with no scaling tier).
+- `skills/al_build_imaging_model.md` triage question pins group at "(2-4)"
+  galaxies — now the multi_galaxy band; merged group/README says 2-10.
+- `wiki/core/external/skill_citation_map.md` — no multi_galaxy row, so a
+  two-deflector request routes to group/start_here.py.
+- `wiki/core/external/index.md` + `external/workspace.md` — stale top-level
+  folder enumerations (mitigated: the regenerated llms-full.txt /
+  workspace_index.json the assistant is told to read DO contain
+  multi_galaxy).
+- Nothing in the assistant references `group/features/group_halo/`.
+
 ## Ordering
 
 Blocked on: multi_galaxy_package, group_halo_explicit_choice,
-cluster_regime_narrative, autogalaxy packages. Do not start before those
-merge — the assistant must document the shipped surface, not the plan.
+cluster_regime_narrative, autogalaxy packages. The workspace/docs core
+shipped 2026-07-25 (autolens_workspace#346 + companions), so the
+autolens_assistant leg is now unblocked; the autogalaxy-side leg still
+waits on the autogalaxy packages.

draft/docs/workspaces/regime_conventions_stragglers.md

@@ -0,0 +1,36 @@
+# Regime-split stragglers: consumers still on pre-split conventions
+
+Type: docs
+Target: autolens_workspace
+Repos:
+- autolens_workspace
+- HowToLens
+Difficulty: small
+Autonomy: supervised
+Priority: normal
+Status: draft
+Parent: draft/docs/autolens/split_lensing_regimes.md
+
+The post-merge Opus residual sweep (2026-07-25) found three consumers still
+on pre-split conventions after the regime reorganization merged. None are
+CI-visible; all need a light touch.
+
+1. `autolens_workspace/scripts/weak/features/strong_lensing/a2744.py`
+   (lines ~177-193) still uses all three superseded cluster conventions the
+   merged guide repudiates: `scaling_radius_exponent = 0.5`,
+   `r_cut_ref = 15.8`, and a luminosity-scaled `r_core = 0.158 * L^0.5` —
+   while its own input `dataset/cluster/a2744/mass.csv` now carries the
+   vanishing-core convention. no_run-skipped (SLOW), so nothing surfaces it.
+   Needs the swept conventions applied + a quick science sanity check that
+   the weak-lensing comparison it feeds is unaffected.
+2. `HowToLens/scripts/simulator/lens_x2.py` (lines ~10-11) routes a
+   two-deflector system to the `group` package — that is now precisely the
+   `multi_galaxy` exemplar; update the pointer (and any surrounding prose).
+3. `autolens_workspace/scripts/cluster/lenstool/parameterization_mapping.py`
+   (lines ~180-191) documents Lenstool's native L^0.5 / 15.8" / 0.158"
+   convention — defensibly, since its job is Lenstool interop — but never
+   labels it as the legacy convention vs the modern tied exponent now used
+   by `scripts/cluster/`. Add one labelling sentence.
+
+Acceptance: smoke suite green; notebooks + navigator regenerated for any
+touched workspace scripts; HowToLens smoke green on its own CI.