FE-846: Web revamp pass II

.agents/skills/ln-review/SKILL.md

@@ -60,6 +60,7 @@ Concrete cues to look for:
 - A magic check inferring readiness/state from an object's incidental shape instead of a named constant or predicate. Repair: name the predicate against the canonical constant.
 - Ordering or position encoded by a numeric index/splice rather than by structure. Repair: make the order declarative.
 - A type alias or name that implies a wider contract than it points at. Repair: point it at the real union, or rename.
+- A method-shaped read/cache surface added without the matching update path (RPC method, query key, publisher topic, client invalidator, and README ledger entry drifting apart). Repair: thread the method-shaped topic through the whole publish/invalidate path, and lock it with a narrow invalidation/publisher test.
 
 Collect findings as numbered items (category: `contract`). Frame each as: the assumed contract in one sentence, the failure mode when it breaks, and which of the three repairs applies. Most are concrete fixes (`ln-scope`/`ln-build`); clusters across a seam route to `ln-refactor`.
 

.agents/skills/ln-scope/SKILL.md

@@ -74,7 +74,7 @@ Created:  YYYY-MM-DD
 
 The `memory/cards/` directory is a scoping inbox where multiple agents can deposit independent scope files in parallel without colliding on a single shared file. Each file is the unit of work one `ln-build` invocation consumes.
 
-The card does **not** inline canonical context — it points to it via [§Cold-start reads](#cold-start-reads). The full execution context is the card *plus* the canonical docs its Cold-start reads enumerate, which `ln-build` reloads on a fresh thread. A card therefore need not be self-contained to be cold-buildable; it must make its required reads explicit. "Free-standing enough for a separate builder thread" means *its Cold-start reads are complete*, not *its content is duplicated* — inlining SPEC/PLAN text into the card duplicates canonical truth and invites drift.
+The card does **not** inline canonical context — it points to the Cold-start reads block in whichever card template it uses. The full execution context is the card *plus* the canonical docs its Cold-start reads enumerate, which `ln-build` reloads on a fresh thread. A card therefore need not be self-contained to be cold-buildable; it must make its required reads explicit. "Free-standing enough for a separate builder thread" means *its Cold-start reads are complete*, not *its content is duplicated* — inlining SPEC/PLAN text into the card duplicates canonical truth and invites drift.
 
 Multiple scope files per frontier are permitted — they represent independent concerns that happen to land on the same branch. They do **not** imply multiple Linear issues or multiple Graphite branches; the frontier item remains the tracker/branch boundary.
 
@@ -183,7 +183,7 @@ If you cannot name the containing seam, the governing decision, or the live inva
 
 What is true when this slice is done? Single declarative sentence — observable, testable, no conjunctions.
 
-### Cold-start reads
+### Full-card cold-start reads
 
 The canonical context a fresh builder thread must resolve **before** building this card. Pointers, not copies — name the exact ids/paths to load; never restate their content here (that duplicates canonical truth and invites drift).
 
@@ -297,7 +297,7 @@ src/legacy/observer.ts ?
 
 Single sentence: what this work changes for the user, operator, or codebase.
 
-### Cold-start reads
+### Light-card cold-start reads
 
 The canonical pointers a fresh builder must resolve before building — ids/paths, not copies.
 

.fixtures/README.md

@@ -1,11 +1,12 @@
 # `.fixtures/`
 
-Current seed data plus probe artifacts and transcript evidence for the Brunch POC.
-The active convention is **probe first, transcript-backed**: each committed run
-must have a probe id, a run id, executable/reportable oracle output, and the
-transcript artifact needed for human review. Brief-based golden fixtures may
-return later, but they should be generated through this probe/transcript path
-rather than a separate brief-library subsystem.
+Current seed data, launchable workbenches, curated probe artifacts, and ephemeral
+dev-loop scratch output for the Brunch POC. The active convention for committed
+evidence is **probe first, transcript-backed**: each committed run must have a
+probe id, a run id, executable/reportable oracle output, and the transcript
+artifact needed for human review. Brief-based golden fixtures may return later,
+but they should be generated through this probe/transcript path rather than a
+separate brief-library subsystem.
 
 See [`docs/architecture/probes-and-transcripts.md`](../docs/architecture/probes-and-transcripts.md)
 for the current architecture.
@@ -14,20 +15,31 @@ for the current architecture.
 
 ```
 .fixtures/
-├── seeds/
+├── seeds/                      # Tracked reusable explicit-basis inputs
 │   └── <seed-set>/
 │       ├── README.md
-│       ├── <seed>.json          # Reusable explicit-basis starting truth
-│       └── _*.ts                # Reproducible data-prep scripts, not product code
-└── runs/
-    └── <probe-id>/
+│       ├── <seed>.json
+│       └── _*.ts               # Reproducible data-prep scripts, not product code
+├── workbenches/                # Launchable local workspaces; .brunch/ is gitignored
+│   └── <name>/
+├── runs/                       # Tracked curated/promoted probe evidence
+│   └── <probe-id>/
+│       └── <run-id>/
+│           ├── session.jsonl        # Source transcript / canonical run evidence
+│           ├── transcript.md        # Human-readable semantic rendering
+│           ├── report.json          # Probe report and artifact paths
+│           └── graph-overview.json  # Optional graph readback when graph truth is the proof target
+└── scratch/                    # Gitignored ephemeral dev-loop output
+    └── <loop>/
         └── <run-id>/
-            ├── session.jsonl        # Source transcript / canonical run evidence
-            ├── transcript.md        # Human-readable semantic rendering
-            ├── report.json          # Probe report and artifact paths
-            └── graph-overview.json  # Optional graph readback when graph truth is the proof target
 ```
 
+Promote scratch to evidence only deliberately: move a reviewed
+`scratch/<loop>/<run-id>/` under `runs/<probe-id>/<run-id>/`, add the missing
+probe report/transcript artifacts, then track it. Dev launchers must resolve
+scratch from the repo-root `.fixtures/scratch/`, independent of the workspace cwd
+they target.
+
 ## Current runs
 
 - `runs/public-rpc-parity/2026-05-29-public-rpc-parity/` — FE-744 public Brunch

.gitignore

@@ -38,6 +38,8 @@ bun.lock
 .brunch/
 brunch.db*
 todo.txt
+.fixtures/scratch/
+.fixtures/workbenches/*/.brunch/
 
 # Claude Code worktrees
 .claude/worktrees/