docs: require impeccable skill for design and frontend

Adds the design-system issue, scaffolds .impeccable/live/config.json against
web/index.html, and records the aoe spawn commands for the codex GPT-5.6
models in the driver table.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.impeccable/live/config.json

@@ -0,0 +1,6 @@
+{
+  "files": ["web/index.html"],
+  "insertBefore": "</body>",
+  "commentSyntax": "html",
+  "cspChecked": false
+}

CLAUDE.md

@@ -42,6 +42,27 @@ request that adds a heavy dependency to either is wrong.
 - No `unwrap()` in non-test code. The lint is on.
 - SQL goes through `sqlx` compile-time-checked queries.
 
+## Design and frontend work
+
+**Any issue labelled `area/web`, and any work that decides how something looks
+or behaves on screen, goes through the `impeccable` skill.** That includes the
+design system itself, component work, layout, visual hierarchy, empty and error
+states, and any change to the UI described in `DESIGN.md` §9.2 and §9.3. Invoke
+it before writing markup, not as a review pass afterwards.
+
+`.impeccable/design.json` is the token system — colours, typography, spacing.
+It is authored once by that skill and then treated as the source of truth;
+components consume tokens rather than literal values. `.impeccable/live/config.json`
+points the skill's live browser iteration at `web/index.html`.
+
+`~/tea/arcada` has the same setup and is worth reading for the shape of a
+finished `design.json`.
+
+This matters more here than in a typical CRUD app because the manual-search
+view (§9.3) is the whole reason for the project's UI existing — Radarr's is
+unusable specifically because of a layout decision. Getting it right is a design
+problem, not a markup problem.
+
 ## Working an issue
 
 1. Read `DESIGN.md`, then the issue, then the sections the issue cites.
@@ -110,19 +131,28 @@ alone. The loop:
 
 ### Model selection
 
-| Difficulty | Model |
-|---|---|
-| `difficulty/hard` | Opus 5 |
-| `difficulty/moderate` | Sonnet 5 |
-| `difficulty/easy` | Sonnet 5 |
-| `difficulty/trivial` | Haiku 4.5 |
+Sessions are spawned with `aoe`, which can run `claude`, `codex`, `opencode` and
+`pi`. The GPT-5.6 models are `terra` and `sol`, reached through the `codex` tool.
+
+| Difficulty | Model | Spawn |
+|---|---|---|
+| `difficulty/hard` | Opus 5 | `aoe add <path> --tool claude --model claude-opus-5 -l` |
+| `difficulty/moderate` | Sonnet 5 or `gpt-5.6-sol` | `aoe add <path> --tool codex --extra-args "-m gpt-5.6-sol" -l` |
+| `difficulty/easy` | `gpt-5.6-terra` | `aoe add <path> --tool codex --extra-args "-m gpt-5.6-terra" -l` |
+| `difficulty/trivial` | Haiku 4.5 | `aoe add <path> --tool claude --model claude-haiku-4-5-20251001 -l` |
+
+Use `-w <branch> -b` to put each session in its own git worktree, which is what
+makes parallel issues safe.
+
+`gpt-5.6-terra` is confirmed — it is the default in `~/.codex/config.toml`.
+`gpt-5.6-sol` is inferred from the sibling naming and should be verified on
+first use.
 
 Escalate one tier if a session fails CI twice on the same issue. Never
 de-escalate mid-issue.
 
-> The operator mentioned a model named "terra" for the lower tiers. No such
-> model was identified — confirm what it refers to and correct this table
-> before relying on it.
+`area/web` issues ignore this table's lower tiers — see the design section
+above. They go through the `impeccable` skill regardless of difficulty label.
 
 ### What a driver must not do
 

scripts/seed-issues.py

@@ -170,6 +170,9 @@ embedded into the binary with `include_dir` and served with correct MIME types.
 - Dev mode proxies `/api/*` to the running daemon so the SPA has hot reload.
 - One page that calls `GET /api/health` and renders it, to prove the whole path.
 
+Use the `impeccable` skill for this — see the design section of `CLAUDE.md`.
+
+
 Acceptance: `cargo run` serves the built SPA from a single binary with no
 external files.
 """,
@@ -625,10 +628,40 @@ client compiles.
     ),
 
     # ---------------------------------------------------------------- phase 5
+    issue(
+        "design", "Design system tokens",
+        "phase/5-ui", "area/web", "difficulty/moderate",
+        deps=["web"],
+        body="""
+Author `.impeccable/design.json` — colours, typography, spacing, the token set
+every component consumes. **Use the `impeccable` skill for this**; it is what
+the file's schema belongs to (`CLAUDE.md`, design section).
+
+`~/tea/arcada` has a finished `design.json` at `schemaVersion: 2` worth reading
+first for the shape.
+
+Constraints specific to this app:
+
+- The manual-search view (`DESIGN.md` §9.3) is dense, tabular and chip-heavy.
+  The palette has to carry three verdict states — eligible, waived, rejected —
+  legibly at small sizes and against each other, without relying on colour
+  alone.
+- Library status (§4.2) is five states, and they are informational rather than
+  severity-ranked. Do not borrow an error/warning/success ramp for them.
+- It runs on a NAS dashboard, often on a phone, often in a dark room. Dark mode
+  is not an afterthought.
+
+`.impeccable/live/config.json` is already scaffolded and points at
+`web/index.html`.
+
+Acceptance: components consume tokens, never literal colour or spacing values,
+and CI can assert that.
+""",
+    ),
     issue(
         "uisearch", "Unified search box",
         "phase/5-ui", "area/web", "difficulty/moderate",
-        deps=["web", "searchapi", "moviesapi"],
+        deps=["web", "design", "searchapi", "moviesapi"],
         body="""
 `DESIGN.md` §9.2. One box, results grouped: in-library first, TMDB below. Enter
 on a TMDB result opens the add flow with root and policy pre-filled.
@@ -639,13 +672,16 @@ The requirement, stated as a test: there is never a moment where the user has to
 know whether they are searching or adding. If the UI has an "Add Movie" page
 separate from search, this issue is not done.
 
+Use the `impeccable` skill for this — see the design section of `CLAUDE.md`.
+
+
 Debounce, and cancel in-flight requests on keystroke.
 """,
     ),
     issue(
         "uimanual", "Manual search buckets and attribute chips",
         "phase/5-ui", "area/web", "difficulty/moderate",
-        deps=["web", "searchapi"],
+        deps=["web", "design", "searchapi"],
         body="""
 `DESIGN.md` §9.3. The fix for Radarr's unusable manual search.
 
@@ -659,12 +695,15 @@ Debounce, and cancel in-flight requests on keystroke.
 - One click on a `waived` row grabs it and writes the override.
 
 The client must not re-implement any policy. Verdicts arrive from the API.
+
+Use the `impeccable` skill for this — see the design section of `CLAUDE.md`.
+
 """,
     ),
     issue(
         "uilibrary", "Library views with derived status",
         "phase/5-ui", "area/web", "difficulty/moderate",
-        deps=["web", "moviesapi"],
+        deps=["web", "design", "moviesapi"],
         body="""
 `DESIGN.md` §4.2. The default view shows `airing` and `incomplete` only;
 everything else collapses behind one toggle.
@@ -675,12 +714,15 @@ the exact problem this design removed.
 
 Show waivers honestly: a file imported under `allow_english_audio` reads as
 "English, no dub", not as a clean match.
+
+Use the `impeccable` skill for this — see the design section of `CLAUDE.md`.
+
 """,
     ),
     issue(
         "uiqueues", "Attention queues: no-PT-source and needs-decision",
         "phase/5-ui", "area/web", "difficulty/easy",
-        deps=["uilibrary", "failure"],
+        deps=["uilibrary", "design", "failure"],
         body="""
 `DESIGN.md` §5.2 and §5.7. Two lists of things a human has to look at.
 
@@ -691,6 +733,9 @@ Show waivers honestly: a file imported under `allow_english_audio` reads as
 
 These are the only places the app asks for attention, and they are the same
 events that notify (§9.5). Keep them consistent.
+
+Use the `impeccable` skill for this — see the design section of `CLAUDE.md`.
+
 """,
     ),
 
@@ -929,32 +974,42 @@ def main():
     by_title = {i["title"]: i["number"] for i in open_issues}
     numbers = {}
 
+    # Pass 1 — ensure every issue exists, so dependency numbers are all known
+    # before any body is rendered.
     for spec in ISSUES:
         title = spec["title"]
         if title in by_title:
             numbers[spec["key"]] = by_title[title]
-            print(f"issue = #{by_title[title]} {title}")
             continue
-
-        body = spec["body"]
-        if spec["deps"]:
-            refs = []
-            for d in spec["deps"]:
-                refs.append(f"#{numbers[d]}" if d in numbers else f"`{d}` (not yet created)")
-            body += "\n\n---\n\n**Depends on:** " + ", ".join(refs)
-        body += f"\n\nLabels: `{'`, `'.join(spec['labels'])}`"
-
         print(f"issue + {title}")
         if DRY:
             continue
         made = req("POST", f"/repos/{OWNER}/{REPO}/issues", {
             "title": title,
-            "body": body,
+            "body": spec["body"],
             "labels": [label_ids[n] for n in spec["labels"] if n in label_ids],
         })
         if made:
             numbers[spec["key"]] = made["number"]
 
+    # Pass 2 — render bodies with real dependency numbers and push them.
+    # Declarative: this overwrites hand-edits to issue bodies. Edit the spec
+    # here, not the issue on Gitea.
+    for spec in ISSUES:
+        if spec["key"] not in numbers:
+            continue
+        body = spec["body"]
+        if spec["deps"]:
+            refs = ", ".join(f"#{numbers[d]}" for d in spec["deps"] if d in numbers)
+            body += f"\n\n---\n\n**Depends on:** {refs}"
+        body += f"\n\nLabels: `{'`, `'.join(spec['labels'])}`"
+        if DRY:
+            continue
+        req("PATCH", f"/repos/{OWNER}/{REPO}/issues/{numbers[spec['key']]}", {
+            "body": body,
+            "labels": [label_ids[n] for n in spec["labels"] if n in label_ids],
+        })
+
     print(f"\n{len(numbers)}/{len(ISSUES)} issues present")