first-fluke · gracefullight · Jul 10, 2026 · Jul 6, 2026
@@ -116,6 +116,7 @@ export function primerContext(): string {
     "- Code discovery / reading: `get_symbols_overview`, `find_symbol`, `find_referencing_symbols`, `search_for_pattern`.",
     "- Code edits: `replace_symbol_body`, `insert_after_symbol`, `insert_before_symbol`, `replace_content`.",
     "- Native grep/glob: only for initial filename/path discovery. Do not fall back to grep + Read for code navigation just because Serena's tools aren't loaded yet — load them.",
+    "- Exception — MCP timeout: if a Serena MCP call times out or hangs (seen mainly in OpenCode Desktop's long-lived sidecar), stop retrying MCP for this session: use native search/read for code, and access `.serena/memories/` files directly (or `serena memories read|write` when Serena CLI ≥ 1.5 is installed) for memory work. A full app relaunch restores Serena MCP.",
   ].join("\n");
 }
 

@@ -2067,7 +2067,8 @@
           "react",
           "tailwind",
           "tsx",
-          "frontend"
+          "frontend",
+          "angular"
         ],
         "en": [
           "make a react component",
@@ -2086,7 +2087,8 @@
           "넥스트 페이지",
           "tailwind로 스타일",
           "shadcn 붙여줘",
-          "프론트 구현"
+          "프론트 구현",
+          "앵귤러"
         ],
         "ja": [
           "reactコンポーネント",

@@ -191,14 +191,14 @@ Unlike CD (tracked in real-time), Evaluator Accuracy (EA) is a
 |-------|--------|-----------------|
 | `false_negative` | +30 | Next session or production: bug that QA missed |
 | `false_positive` | +15 | During session: impl agent disputes QA finding successfully |
-| `severity_mismatch` | +10 | During session or retro: wrong severity assigned |
+| `severity_mismatch` | +10 | During session or next-session review: wrong severity assigned |
 | `missed_stub` | +20 | During session: runtime verification catches display-only feature |
 | `good_catch` | -10 | During session: QA caught non-obvious bug (reward signal) |
 
 ### Recording
 
 - `false_positive`, `missed_stub`, `good_catch`: Recorded during session by Orchestrator
-- `false_negative`, `severity_mismatch`: Recorded retroactively via `oma retro` or next session discovery
+- `false_negative`, `severity_mismatch`: Recorded retroactively on next-session discovery
 
 ### Evaluator Accuracy Score (EA)
 
@@ -210,10 +210,10 @@ EA = sum(accuracy_event_points across last 3 sessions)
 
 | Threshold | Action |
 |-----------|--------|
-| EA >= 30 | **TUNING SUGGESTED**: `oma retro` flags QA patterns for review |
+| EA >= 30 | **TUNING SUGGESTED**: Review accumulated EA events for recurring QA judgment errors |
 | EA >= 50 | **TUNING REQUIRED**: Review and update QA execution-protocol.md |
 | `false_negative` >= 3 across window | **CHECKLIST UPDATE**: Add detection pattern to QA checklist.md |
-| `good_catch` >= 5 across window | **PROPAGATE**: Document successful pattern in evaluator-tuning.md |
+| `good_catch` >= 5 across window | **PROPAGATE**: Generalize the successful pattern into `common-checklist.md` |
 
 ### Accuracy Log Format
 

@@ -15,6 +15,25 @@ If Serena MCP is available, use `read_memory`/`write_memory`/`edit_memory` (its
 base path is `.serena/memories`); otherwise write the same files there directly with
 opencode's file tools.
 
+### Serena MCP Timeout Recovery (OpenCode Desktop)
+
+OpenCode Desktop runs one long-lived sidecar server; new sessions reuse its MCP clients,
+so a stuck Serena MCP stays stuck until the Desktop app is fully relaunched (the TUI is
+rarely affected). When a Serena MCP call times out or the MCP queue is clearly stuck, do
+not keep retrying MCP — fall back narrowly:
+
+1. **Memory ops** — read/write the same files directly under `.serena/memories/` (as
+   above). If a recent Serena CLI is installed (≥ 1.5, check `serena --version`),
+   `serena memories read|write|list` is an equivalent alternative; older versions
+   (e.g. 1.3.x) do not have the `memories` command.
+2. **Code analysis** — fall back to native search/read tools. The Serena CLI cannot
+   execute analysis tools (`serena tools` only lists/describes them).
+3. **Diagnostics** — `serena project health-check` and `serena project index` work
+   without MCP.
+
+Keep the fallback scoped to the blocked call: this is a recovery path, not a license to
+abandon Serena-first. A full Desktop relaunch is what actually resets the stale MCP client.
+
 ### Path Resolution (CRITICAL)
 
 All result, progress, and state files MUST be written to the **project root** `.serena/memories/` directory, never to a subdirectory's `.serena/memories/`.

@@ -1,6 +1,6 @@
 {
-  "version": "10.7.1",
+  "version": "10.10.0",
   "schemaVersion": 2,
   "mode": "project",
-  "installedAt": "2026-06-28T01:08:28.353Z"
+  "installedAt": "2026-07-06T12:39:22.805Z"
 }
@@ -24,6 +24,7 @@ DESIGN.md is the central artifact; all design work revolves around it.
 - Creating or auditing `DESIGN.md`
 - Selecting typography, color, layout, motion, or component direction
 - Reviewing UI work for responsive behavior, accessibility, and visual quality
+- Redesigning an existing site or app (preserve vs overhaul modes)
 - Using optional vendor inspiration from Stitch MCP or getdesign
 
 ### When NOT to use
@@ -81,6 +82,7 @@ outputs:
 
 ### Transitions
 - If `.design-context.md` is missing, create it before continuing.
+- If the target is an existing site/app, load `resources/redesign-protocol.md` and classify Preserve vs Overhaul before proposing.
 - If CJK support is needed, prioritize CJK-ready fonts.
 - If vendor seed fetch fails, choose retry, continue without seed, or abort.
 - If anti-patterns appear, surface alternatives before finalizing.
@@ -151,12 +153,16 @@ bunx getdesign@latest list
 8. WCAG AA minimum for all designs. Respect `prefers-reduced-motion`.
 9. Stitch MCP is optional; all phases work without it.
 10. Present 2-3 design directions and get user confirmation before generating.
+11. Declare a one-line Design Read before proposing: "Reading this as: \<page kind> for \<audience>, with a \<vibe> language." If genuinely ambiguous, ask exactly ONE clarifying question — never guess, never a question dump.
+12. Redesigns follow `resources/redesign-protocol.md`: detect Preserve vs Overhaul, audit before touching, never silently change URLs, nav labels, form field names, or brand marks.
+13. Visual assets follow `resources/asset-strategy.md`: image generation (oma-image) first, picsum seed second, labeled placeholder last. Div-based fake screenshots are banned.
+14. Consistency locks: one accent color, one corner-radius system, one theme per page. Lock them early, audit against them in Phase 6 (checklist section 6 mechanical checks).
 
 ### Anti-Pattern Quick Reference
 
 ### Typography
 - DON'T: Default to custom Google Fonts when system fonts suffice
-- DON'T: Use Inter/Geist alone without considering project context
+- DON'T: Reach for Inter as the default custom sans (LLM signature); override only on explicit neutral/Linear-style ask or public-sector/a11y-first brief
 - DON'T: Load 3+ font families without justification
 - DON'T: Body text below 16px on mobile
 - DO: System font stack first, custom fonts for brand identity only
@@ -190,9 +196,21 @@ bunx getdesign@latest list
 ### Components
 - DON'T: Glassmorphism everywhere; use sparingly
 - DON'T: Hover-only interactions without touch/keyboard alternatives
+- DON'T: Default to lucide-react or hand-rolled SVG icons; pick one family deliberately (Phosphor > HugeIcons > Radix > Tabler; lucide on explicit ask or existing dependency)
 - DO: shadcn/ui for base, Aceternity UI / React Bits for accent effects
 - DO: All interactive elements must have visible focus states
 
+### Content & Copy
+- DON'T: "John Doe" personas, "Acme/Nexus" brand names, Lorem Ipsum, fake-perfect numbers (99.99%)
+- DON'T: Filler verbs ("Elevate", "Seamless", "Unleash") or performative labels ("Quietly trusted by")
+- DON'T: Em-dash (`—`) anywhere in visible output — zero tolerance, no "sparingly" allowance; en-dash as separator also banned (hyphen only)
+- DO: Copy self-audit before handoff — re-read every visible string, rewrite anything broken or AI-cute
+- DO: One label per CTA intent per page
+
+### Assets
+- DON'T: Div-based fake screenshots; text wordmark logo walls; hand-rolled SVG icons
+- DO: oma-image generation first → picsum seed → labeled placeholder + report
+
 ### Workflow Summary
 7 phases: Setup → Extract → Enhance → Propose → Generate → Audit → Handoff.
 See `resources/execution-protocol.md` for full detail.
@@ -235,12 +253,14 @@ injection defenses, and multi-vendor merge policy live in
 ### Resources
 - `resources/execution-protocol.md`: 7-phase workflow
 - `resources/anti-patterns.md`: Full DO/DON'T catalog
-- `resources/checklist.md`: Audit checklist (Responsive + WCAG + Nielsen + Slop)
+- `resources/checklist.md`: Audit checklist (Responsive + WCAG + Nielsen + Slop + Mechanical counts)
 - `resources/design-md-spec.md`: DESIGN.md generation guide (9 sections)
 - `resources/design-tokens.md`: CSS/Tailwind/shadcn export templates
 - `resources/prompt-enhancement.md`: Vague request to detailed spec
 - `resources/stitch-integration.md`: Stitch MCP tool mapping (optional)
 - `resources/getdesign-fetcher.md`: Vendor seed fetch, hash verify, seed rules
+- `resources/redesign-protocol.md`: Preserve/Overhaul modes, audit-first, modernization levers
+- `resources/asset-strategy.md`: Image sourcing priority, logo walls, fake-screenshot ban
 - `resources/error-playbook.md`: Design error recovery
 
 ## References

@@ -5,7 +5,7 @@
 
 ## Typography
 - DON'T: Default to custom Google Fonts when system fonts suffice
-- DON'T: Use Inter or Geist alone without considering project context
+- DON'T: Reach for Inter as the default custom sans. When a custom font is justified (brand identity), pick deliberately from alternatives first (Geist, Outfit, Cabinet Grotesk, Satoshi, or a brand-appropriate face). Inter-by-reflex is the LLM signature. OVERRIDE: Inter is acceptable when the user explicitly asks for a neutral / standard / Linear-style feel, or the brief is public-sector / accessibility-first. CJK rules always win (Pretendard Variable / Noto Sans CJK)
 - DON'T: Load 3+ font families without clear justification
 - DON'T: Set body text below 16px on mobile
 - DON'T: Use light font-weight (300) for body on dark backgrounds without testing contrast
@@ -63,6 +63,47 @@
 - DO: Use Intersection Observer to trigger animations only when visible
 - DO: Pause off-screen Canvas/WebGL renderers
 
+## Content & Copy
+- DON'T: Generic persona names ("John Doe", "Jane Smith") or slop brand names ("Acme", "Nexus", "SmartFlow", "Cloudly")
+- DON'T: Fake-perfect numbers (99.99%, 50%, 1234567); use organic data (47.2%) or label mocks explicitly
+- DON'T: Fake-precise specs the brand never claimed (invented "5.8mm", "4.1x" for spec aesthetics)
+- DON'T: Filler verbs — "Elevate", "Seamless", "Unleash", "Next-Gen", "Revolutionize", "Game-changer"
+- DON'T: Lorem Ipsum; write real draft copy
+- DON'T: Poetic/performative section labels ("From the field", "Quietly trusted by", "On our desks"); use plain functional labels ("Testimonials", "Latest writing") or none
+- DON'T: Section-number eyebrows ("001 · Capabilities", "06 · how it works") or version labels in the hero (BETA, V0.6) unless the brief is a launch
+- DON'T: Scroll cues ("Scroll to explore", animated mouse icons) or decorative locale/time/weather strips
+- DON'T: Mix copy registers (technical mono + editorial prose + marketing punch) in one page without brand justification
+- DON'T: Quotes longer than 3 lines; attribution is name + role, never name only ("- Sarah")
+- DON'T: Em-dash (`—`) ANYWHERE in visible output — the #1 LLM stylistic tell. Zero tolerance: banned in headlines, eyebrows, pills, body copy, quotes, attribution, captions, button text, and alt text. No "sparingly" allowance — the model ignores soft limits. Restructure instead: two sentences, a comma, parentheses, or a colon
+- DON'T: En-dash (`–`) as a separator; date ranges (2018-2026) and number ranges use a plain hyphen. The only permitted dashes are the hyphen (`-`) and the math minus sign
+- DO: Copy self-audit before handoff — re-read every visible string; rewrite anything grammatically broken, unclear-referent, or AI-cute. Plain functional copy beats clever broken copy
+- DO: One label per CTA intent across the page
+- DO: Apply register rules per content language (CJK copy follows i18n-guide)
+
+## Consistency Locks
+- DON'T: Introduce a second accent color mid-page (a rose-accented site does not get a teal badge in the footer)
+- DON'T: Mix corner-radius systems without a documented rule (round buttons in a square layout is broken design)
+- DON'T: Invert theme for a single section mid-page (one light warm-paper section inside a dark page reads as a paste accident)
+- DO: Pick accent, radius scale, and theme ONCE per page; lock them; audit every component against the locks before handoff
+- DO: If a deliberate full theme switch is a composition device, use it at most once per page with a strong transition
+
+## Assets & Imagery
+- DON'T: Div-based fake screenshots (fake dashboards, task lists, terminal windows built from styled divs) — the #1 AI-design tell
+- DON'T: Text-only pages passing as "minimalism"; even restrained pages need 2-3 real images
+- DON'T: Plain text wordmarks in "Trusted by" logo walls; use real SVG logos or generated monogram marks
+- DON'T: Hand-rolled SVG icons or decorative illustrations by default
+- DON'T: Pills/labels overlaid on images, photo-credit captions as decoration, guessed Unsplash URLs
+- DO: Source priority: image generation (oma-image) → picsum seed → labeled placeholder slot + tell the user
+- DO: See `resources/asset-strategy.md` for full rules
+
+## Iconography
+- DON'T: Default to `lucide-react`. It ships with shadcn/ui, so every AI-generated site carries the same icon set. OVERRIDE: acceptable when the user explicitly asks for it, or the project already depends on it (including shadcn base components — never mix families to escape it)
+- DON'T: Hand-roll SVG icon paths. If a glyph is missing, install a second allowed library or compose from primitives
+- DON'T: Mix icon families in one component tree (no Phosphor + Lucide together)
+- DON'T: Inconsistent stroke widths; standardize strokeWidth globally (e.g. 1.5 or 2.0)
+- DO: Pick deliberately, priority order: `@phosphor-icons/react` > `hugeicons-react` > `@radix-ui/react-icons` > `@tabler/icons-react`
+- DO: One icon family per project, declared in DESIGN.md
+
 ## Components
 - DON'T: Glassmorphism on every element; reserve for badges, nav pills, accent cards
 - DON'T: Icon + Title + Description card grid as the only layout pattern