lazycodex-ai 5.0.0-beta.9 → 5.0.0

package/packages/shared-skills/skills/debugging/references/tools/dap.md

@@ -0,0 +1,103 @@
+# DAP Client (`dap.mjs`) — Drive the Debugger's Protocol, Not Its Text
+
+**Design modeled on the `debug` tool of oh-my-pi (https://github.com/can1357/oh-my-pi, branch omp2).** Their harness proved the shape: one structured debug surface with bounded output, stop snapshots, and classified errors. This script brings the same discipline to any agent with a shell — no harness tool registration required.
+
+Debuggers already speak the Debug Adapter Protocol (DAP), a machine-readable JSON protocol. Driving DAP beats screen-scraping a PTY for the same reason an API beats OCR: structured stops, structured variables, structured errors. **If the debugger speaks DAP, use this script instead of parsing `gdb`/`pdb` output.**
+
+The script is at `references/scripts/dap.mjs` — zero dependencies, runs under Bun or Node.
+
+---
+
+## When to use which interface
+
+| Situation | Use |
+|---|---|
+| Source-level debugging of Python / Go / Node / native code, and you need breakpoints, stepping, variables | `dap.mjs` (this file) |
+| Browser-served JS, or anything already in Chrome | Scripted CDP — see `references/runtimes/node.md` |
+| Stripped binary, no source, no symbols | Ghidra (static) + Frida (live) — see `references/tools/` |
+| The debugger has no DAP mode (plain gdb on an exotic target) | pwndbg, with the output-budget rule from `references/methodology/00-setup.md` |
+
+---
+
+## How to drive it
+
+`dap.mjs` is a **REPL**, not a one-shot command: a debug session is long-lived, so the script runs as a persistent process reading one command per line on stdin and writing bounded text on stdout. Run it as a background shell session, send commands to its stdin, and subscribe to its output.
+
+```
+bun references/scripts/dap.mjs
+```
+
+Every execution-control answer prints a single snapshot line beginning with `STOP:` — subscribe your watcher to that exact prefix so a breakpoint hit wakes you instead of you polling.
+
+### Commands
+
+| Command | Effect |
+|---|---|
+| `launch <adapter> <program> [args...]` | Spawn a stdio adapter (path to executable or `.mjs` adapter) and launch the program under it |
+| `attach <host:port>` | Connect to a listening adapter over TCP and attach |
+| `break <file>:<line>` / `rmbreak <file>:<line>` | Set / remove a source breakpoint |
+| `continue` / `step` / `next` / `stepin` / `stepout` / `pause` | Execution control; each prints a `STOP:` snapshot when the debuggee next stops |
+| `stack [limit]` | TSV backtrace, bounded |
+| `scopes` | Scopes of the top frame with their variablesReferences |
+| `vars <ref>` | Variables under a variablesReference from `scopes` — refs are session-scoped; always chain `scopes` first |
+| `eval <expr>` | Evaluate in the top frame |
+| `threads` / `sessions` | Thread list / session state |
+| `terminate` / `quit` | End the debuggee / exit the REPL |
+
+### Output contract (what keeps your context alive)
+
+- Tabular results are TSV with a header row.
+- Hard caps: `MAX_ROWS = 100`, `MAX_OUTPUT_BYTES = 32 KB`. Overflow prints an explicit `TRUNCATED: rows dropped=N bytes dropped=M` line — never a silent cut.
+- After every continue/step: `STOP: stopped reason=<why> threadId=<id> <frame> at <file>:<line>:<col>`. On debuggee exit: `EXIT: terminated`.
+- Errors are one line, classified: `ERR: invalid-args | no-session | adapter-failed | unverified-breakpoint | timeout | terminated | adapter-error`. The token tells you the recovery — see the failure taxonomy in `references/methodology/02-investigate.md`.
+- Every request times out after 15 s (override with `DAP_TIMEOUT_MS`) and the session stays alive.
+- `DAP_DEBUG=1` logs raw protocol traffic to stderr for diagnosing a misbehaving adapter.
+
+### Adapter quirks this client already handles
+
+- **debugpy** requires `console: "internalConsole"` in the launch request, and withholds the `initialized` event for unrecognized `adapterID`s — the client sends both correctly. It also flushes `initialized` lazily, so the client launches first and awaits the event after.
+- **lldb-dap** reports the launch response's `success` unreliably (`false` even when the process started and stopped) and reports `threadId: 0` on the entry stop. The events are the real signal; the client tolerates the response flag.
+
+---
+
+## Adapter matrix
+
+Translated from oh-my-pi's `builtin_adapters()` (omp2 `crates/docserver/src/dap_adapter.rs`). Install the one matching your runtime; the client speaks to all of them the same way.
+
+| Runtime | Adapter | Install | Launch mode |
+|---|---|---|---|
+| Python | debugpy | `pip install debugpy` | `python -m debugpy.adapter` (stdio) |
+| Go | dlv | `go install github.com/go-delve/delve/cmd/dlv@latest` | `dlv dap` (stdio) |
+| C/C++/Swift/Rust/Zig | lldb-dap | ships with Xcode CLT / LLVM | `lldb-dap` (stdio) |
+| C/C++/Rust/Zig | codelldb | VS Code extension binary | stdio |
+| C/C++/Rust | gdb | `brew install gdb` / apt | via gdb's DAP mode where available |
+| JS/TS | js-debug | `npm i -g @vscode/js-debug` | stdio |
+| .NET | netcoredbg | GitHub releases | stdio |
+| Ruby | rdbg | `gem install debug` | stdio |
+| Kotlin | kotlin-debug-adapter | GitHub releases | stdio |
+| PHP | php-debug-adapter | composer | stdio |
+| Bash | bash-debug-adapter | GitHub releases | stdio |
+| Dart/Flutter | dart debug adapter | ships with Dart SDK | `dart` (stdio) |
+| Elixir | elixir-ls debugger | GitHub releases | stdio |
+
+If an adapter is missing, the install line is the fix — do not fall back to scraping a REPL.
+
+---
+
+## Worked example (Python)
+
+```
+launch /path/to/python -m debugpy.adapter myscript.py   # or a wrapper script path
+break myscript.py:12
+continue
+STOP: stopped reason=breakpoint threadId=1 compute at myscript.py:12:1
+stack
+scopes
+vars 6          # the variablesReference scopes just printed for Locals
+eval total
+continue
+terminate
+quit
+```
+
+Cleanup is part of the session: `terminate` the debuggee, `quit` the REPL, and journal any wrapper scripts you created per the skill's cleanup phase.

package/packages/shared-skills/skills/debugging/references/tools/frida.md

@@ -0,0 +1,193 @@
+# Frida — Hook a Live Process and See Real Values at Runtime
+
+**https://frida.re**
+
+Frida injects a JavaScript engine into a running process and lets you intercept function calls, read arguments, patch return values, and trace execution without touching the binary on disk. **When static analysis in Ghidra shows you *what* a function does, Frida shows you what it does *right now*, with real inputs.** The two compose naturally: use Ghidra to find the target function and its offset, then use Frida to hook it and watch live traffic.
+
+Choose Frida over Ghidra when you need runtime values (actual arguments, return codes, heap contents) or when the binary's control flow is too obfuscated for static analysis to resolve.
+
+---
+
+## Install
+
+```bash
+pip install frida-tools            # Python 3 required
+frida --version                    # verify
+```
+
+On macOS, attaching to another user's process needs root (`sudo frida ...`). SIP-protected system binaries (`/usr/bin/*`, `/usr/sbin/*`) are off-limits entirely. On Linux, you may need `ptrace` permissions: run as root or set `kernel.yama.ptrace_scope=0`.
+
+---
+
+## Target discovery with `frida-ps`
+
+```bash
+frida-ps                         # list local processes
+frida-ps -U                      # USB-connected device (iOS/Android)
+frida-ps | grep -i target        # find your target by name
+```
+
+---
+
+## `frida-trace` — zero-code entry point
+
+`frida-trace` auto-generates JavaScript handler stubs for every matched function. You don't write any code to start.
+
+```bash
+frida-trace -p <pid> -i "recv*"                               # trace exported functions by glob
+frida-trace -p <pid> -m "-[NSURLSession dataTaskWithRequest:*]"  # Objective-C method
+frida-trace -U -p <pid> -j "com.example.App!login*"             # Java method (Android)
+frida-trace -f ./target -i "open"                               # spawn + trace from start
+```
+
+Generated handler files land in `__handlers__/`. Each looks like:
+
+```js
+// __handlers__/libSystem.B.dylib/open.js
+{
+  onEnter(log, args, state) {
+    // args[0] is the first argument (a pointer)
+    log(`open("${args[0].readUtf8String()}")`);
+  },
+  onLeave(log, retval, state) {
+    log(`  => ${retval}`);
+  }
+}
+```
+
+Edit that file, save, and `frida-trace` hot-reloads it. This is the fastest path from "I wonder what this function receives" to a concrete answer.
+
+---
+
+## The JS agent API
+
+When `frida-trace` isn't enough, you write a JS agent and load it with `frida` or from a Python host script. These are the building blocks.
+
+### Find a function
+
+```js
+// By exported symbol name
+const openPtr = Module.getExportByName(null, "open");  // null = any module
+const sslWritePtr = Module.getExportByName("libssl.so", "SSL_write");
+
+// By offset in a stripped binary (get the offset from Ghidra)
+const base = Module.getBaseAddress("target");
+const funcPtr = base.add(0x1a3c);
+```
+
+### Intercept calls
+
+```js
+Interceptor.attach(openPtr, {
+  onEnter(args) {
+    // args[0], args[1], ... are NativePointer objects
+    const path = args[0].readUtf8String();
+    const flags = args[1].toInt32();
+    send({ event: "open", path: path, flags: flags });
+  },
+  onLeave(retval) {
+    // retval is the return value as a NativePointer
+    send({ event: "open_ret", fd: retval.toInt32() });
+  }
+});
+```
+
+### Read memory
+
+```js
+args[0].readUtf8String()          // read a UTF-8 C string from a pointer
+args[0].readCString()             // read a raw C string (stops at null)
+args[1].readByteArray(len)        // read len raw bytes (returns ArrayBuffer)
+```
+
+### Replace / stub a function
+
+```js
+// Force a function to always return 0 (e.g. bypass a license check)
+Interceptor.replace(checkLicensePtr, new NativeCallback(function () {
+  return 0;
+}, 'int', []));
+```
+
+Use `Interceptor.replace` sparingly. It changes program behavior, so always note in your journal that you patched something and why.
+
+---
+
+## Python host script
+
+A host script lets you load a JS agent, receive `send()` messages, and post-process results from the Python side.
+
+```python
+import frida, sys
+
+with open("agent.js") as f:
+    agent_source = f.read()
+
+def on_message(message, data):
+    if message["type"] == "send":
+        print(f"[*] {message['payload']}")
+    else:
+        print(f"[!] {message}")
+
+session = frida.attach(int(sys.argv[1]))
+script = session.create_script(agent_source)
+script.on("message", on_message)
+script.load()
+
+try:
+    sys.stdin.read()                 # block until Ctrl+C
+except KeyboardInterrupt:
+    session.detach()
+```
+
+The agent runs inside the target; the host script stays in your terminal and prints structured output.
+
+---
+
+## Output-budget discipline
+
+**Hooking a hot function (e.g. `malloc`, `read`) will flood your context with thousands of lines.** Don't print every call. Aggregate in the agent and report a summary.
+
+```js
+// Count calls, report every 1000
+let callCount = 0;
+Interceptor.attach(mallocPtr, {
+  onEnter(args) {
+    callCount++;
+    if (callCount % 1000 === 0) {
+      send({ event: "malloc_summary", total_calls: callCount });
+    }
+  }
+});
+```
+
+Alternatives: filter by argument value, only log calls from a specific caller module, or batch into an array and `send()` on a timer.
+
+---
+
+## ⚠️ Gotchas
+
+**Attach vs. spawn.** `frida -p <pid>` attaches to an already-running process. If you need to catch early initialization (constructors, `main` entry), spawn the process with `frida -f ./target`. Frida pauses it at startup. Resume with `%resume` in the REPL or by calling `device.resume(pid)` from Python.
+
+**Stripped binaries have no symbol names.** `Module.getExportByName` returns null for internal functions. You must find the function offset in Ghidra, then compute the address at runtime:
+
+```js
+const addr = Module.getBaseAddress("target").add(0x4a20);
+Interceptor.attach(addr, { /* ... */ });
+```
+
+**A crashing target kills your session.** If the target segfaults or aborts, Frida's agent dies with it. Script a respawn loop on the Python side if you expect crashes.
+
+**Thread safety.** `onEnter`/`onLeave` fire on multiple threads. Keep shared state to simple counters.
+
+---
+
+## Cleanup
+
+```bash
+kill <pid>                         # kill any process you spawned with -f
+rm -rf __handlers__/               # remove generated handler stubs
+rm -f agent.js host.py             # remove session-specific scripts
+```
+
+Always note in your journal which processes you attached to and whether you replaced any functions. A forgotten `Interceptor.replace` in a long-running target will cause confusing behavior later.

package/packages/shared-skills/skills/frontend/ATTRIBUTION.md

@@ -217,6 +217,16 @@ documentation text, or component implementations are vendored. beui.dev names re
 property of their owner, and the guide instructs reading component source for mechanism
 extraction only, not vendoring it into references or projects.
 
+`frontend/references/design/ambience-skill.md` is a project-original, curl-only operating
+guide for consulting the react-bits (reactbits.dev) animated-component registry during hero
+atmosphere, background, typographic-motion, and card-surface work, written from live endpoint
+verification; no react-bits source code, documentation text, or component implementations are
+vendored. react-bits is distributed under MIT plus the Commons Clause, which permits use inside
+applications but forbids redistributing or porting the components themselves; the guide
+therefore instructs reading component source for mechanism extraction only and never vendoring
+it into references, design documents, or this repository. React Bits names remain the property
+of their owner.
+
 `frontend/references/design/clone-from-url.md` is a project-original runtime-extraction
 workflow guide. Its browser + `getComputedStyle` clone approach follows the same
 MIT-licensed `JCodesMore/ai-website-cloner-template` clone-website workflow that `aside.md`

package/packages/shared-skills/skills/frontend/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: frontend
-description: "MUST USE for frontend/web UI/UX/visual work: building, styling, redesigning pages/components, React setup, performance audits, visual QA, taste, and polish. Routes four rulesets: design taste router and brand references; perfection for Playwright/Chromium Lighthouse/Core Web Vitals; ui-ux-db palettes/fonts/guidelines; designpowers personas/accessibility/critique/handoff; plus curl-only lazyweb real-app-screen research and the beui.dev interaction catalog. Triggers: frontend, UI, UX, design, redesign, styling, layout, animation, motion, interaction, micro-interaction, make it feel alive, premium, luxury, minimal, brutalist, Awwwards, DESIGN.md, mockup, React, Lighthouse, accessibility, WCAG, Core Web Vitals, looks generic, make it pretty, like X brand, lazyweb, design research."
+description: "Builds, styles, and polishes web UI and UX. Use for any frontend, page, component, styling, layout, animation, or visual-quality task, or when asked to make an interface look or feel a certain way."
 ---
 
 # Frontend
@@ -15,7 +15,9 @@ This file is a router, not a rulebook. The rules live in four rulesets under `re
 |---|---|
 | ANY UI implementation, styling, redesign, mockup, or visual decision | `references/design/README.md` FIRST. It enforces two mandatory gates — the Design System Gate (a `DESIGN.md` must exist before any component is written) and the React Dev Tooling Gate (react-grab / react-scan / react-doctor installed by default) — then routes to the taste and brand references below. |
 | Spatial structure — app shells, scroll ownership, "what goes where", "this layout breaks at X" | ALSO `references/design/layout-skill.md` for the mechanics, then `references/design/stylegallery.md` to fetch a named pattern contract for that exact spatial problem. Both stack on the style skill and add no visual direction. |
+| Paged output — a PDF report, a print stylesheet, a headless-Chrome print pipeline, an HTML deck printed to paper, or any "why is this page half empty" break defect | ALSO `references/design/print-paged-media.md` for the page box, the fragmentation properties, the atomic-block set, and the keep-together side effect that strands a block on a near-empty page. It adds no visual direction and stacks on the routed style skill. |
 | Interaction or motion work — micro-interactions, animated components, transitions, gestures, hover/press/state feedback, "make it feel alive" | ALSO `references/design/interaction-skill.md`. The beui.dev catalog is the mandatory interaction reference: find the nearest pattern, read its real source through the file's curl recipe, and adapt the mechanism to `DESIGN.md` motion tokens. It stacks on the routed style skill — never replaces it. |
+| Hero atmosphere or ambience — an animated, shader, particle, or grid background; a typographic reveal (split, blur, shimmer, typewriter, count-up, marquee); scroll storytelling; spotlight / tilt / glare / glowing-border card surfaces; "make the hero stunning", "add a living background", "animate the headline" | ALSO `references/design/ambience-skill.md`. The react-bits catalog is the mandatory ambience reference: find the nearest pattern in its intent map, read its real source through the file's curl recipe, extract the mechanism, and run its retrofit checklist (reduced motion, off-screen pause, compositor-only, tokens, budget) before shipping. It stacks on the routed style skill — never replaces it, and never vendors react-bits code. |
 | Writing or modifying frontend code, OR auditing performance / SEO / accessibility / quality | ALSO `references/perfection/README.md`. Lighthouse 100 in every category, measured on real Playwright Chromium (never the `lighthouse` CLI), achieved through architecture — never by dropping animations or hiding content. |
 | Looking up a concrete style, palette, font pairing, chart type, landing structure, or UX guideline — or generating a design system from keywords | `references/ui-ux-db/README.md`. A searchable CSV database with a CLI: a lookup tool, not a posture. `design` stays the source of truth for taste and the `DESIGN.md` contract. |
 | ANY implementation or redesign that creates or updates `DESIGN.md` — plus explicit operating-layer asks (personas, critique, debt, handoff, synthetic user testing) | `references/designpowers/README.md` + `lane-c-review.md`. lane-c is the Phase Final flatness/critique reviewer and fills the accessibility-constraint and accepted-debt sections `DESIGN.md` requires. Load other lanes only when their phase applies. |
@@ -65,6 +67,7 @@ The reference library has one architecture file, 12 taste skills (Layer A — *h
 | `output-skill.md` | Stacks on any style skill when output is incomplete — placeholders, `// TODO`, half-done components. |
 | `stitch-skill.md` | Stacks on any style skill for Google Stitch compatibility or a `DESIGN.md` doc export. A complete worked export ships as `stitch-design-example.md`. |
 | `interaction-skill.md` | Stacks on any style skill when work adds or changes interaction or motion. beui.dev-anchored: read the mapped component's source before designing an interaction; reduced motion always. |
+| `ambience-skill.md` | Stacks on any style skill when work adds a hero atmosphere, animated background, typographic reveal, scroll chapter, or card surface effect. react-bits-anchored: read the mapped component's source, extract the mechanism, run the retrofit checklist; one atmosphere per hero, cursor effects quarantined. |
 | `imagegen-frontend-web.md` / `imagegen-frontend-mobile.md` / `imagegen-brandkit.md` | Image-only output (mockup, app-screen concepts, brand board). These NEVER write code — switch to `image-to-code-skill.md` if code is wanted. |
 
 ### Layer B — brand design systems (orthogonal to Layer A; stack freely)
@@ -106,7 +109,7 @@ Domains: `product` `style` `typography` `color` `landing` `chart` `ux` `react` `
 
 ## Ruleset 4 — designpowers (`references/designpowers/`)
 
-`README.md` routes the pinned `Owl-Listener/designpowers` corpus into this workflow. It supplies design context — personas, accessibility and cognitive constraints, critique, debt, handoff, synthetic user testing, motion, role prompts — that must be distilled into `DESIGN.md` first, then used as the implementation contract. It replaces nothing: not this skill, not `/visual-qa`, `/ulw-plan`, `/start-work`, or `/review-work`.
+`README.md` routes the pinned `Owl-Listener/designpowers` corpus into this workflow. It supplies design context — personas, accessibility and cognitive constraints, critique, debt, handoff, synthetic user testing, motion, role prompts — that must be distilled into `DESIGN.md` first, then used as the implementation contract. It replaces nothing: not this skill, not `/visual-qa`, `/ulw-plan`, `/ulw-execute`, or `/review-work`.
 
 ## Quick routes — most common requests
 
@@ -118,6 +121,7 @@ Domains: `product` `style` `typography` `color` `landing` `chart` `ux` `react` `
 | "Premium SaaS hero like Stripe" | `design/README.md` + `design/stripe.md` + `design/soft-skill.md` + `perfection/README.md` |
 | "Improve this existing dashboard" | `design/README.md` + `design/redesign-skill.md` + `perfection/README.md` |
 | "Add micro-interactions" / "animate this" / "make it feel alive" / "polish the interactions" | `design/README.md` + `design/interaction-skill.md` on top of the current style skill + `perfection/README.md` |
+| "Make the hero stunning" / "add an animated background" / "animate the headline" / "count-up / typewriter / marquee" / "spotlight or tilt cards" | `design/README.md` + `design/ambience-skill.md` on top of the current style skill + `perfection/README.md` |
 | "Build this screenshot / Imagen mock / Stitch output exactly" | `design/README.md` + `design/image-to-code-skill.md` + `perfection/README.md` + `/visual-qa` reference-fidelity mode |
 | "Audit my site" / "make this page faster" | `perfection/README.md` (+ `perfection/react-perf-tooling.md` if React) |
 | "Mockup image of a fintech app" — no code | `design/imagegen-frontend-mobile.md` (+ a Layer B brand if named) |
@@ -133,6 +137,7 @@ Domains: `product` `style` `typography` `color` `landing` `chart` `ux` `react` `
 - **Concrete reference = contract.** When a screenshot, mockup, or annotated reference exists, match its pixels, copy, component structure, and responsive intent unless the user accepts a deviation.
 - **Never weaken UX OR flatten the surface to buy points.** No dropping animations, hiding content, simplifying interactions, or replacing rendered/lit material with flat fills and flat geometric primitives for a Lighthouse score or a deadline. Hit 100 AND keep the surface dimensional — both, or neither.
 - **No emojis as icons.** SVG icon sets only (Lucide, Heroicons, Radix, Phosphor).
+- **No coloured accent borders on rounded surfaces.** A `border-l-2 border-primary` stripe on a selected row, a primary-tinted outline on a focused card — any `border-{side}-{primary|warning|destructive|success}` or accent-width rule used to mark selected/focused/active is the most recognizable AI-slop tell in shipped UI. Encode state the way `DESIGN.md` systems do: one ink at many alphas (hover/selected/active wash ramps), a glyph (check) for selection, tonal layering for focus. Keyboard `focus-visible` rings are the only coloured edge allowed. Applies to code you write AND to pre-existing instances on any surface you touch — sweep them out.
 - **GPU-composited animation only** — `transform`, `opacity`, `filter`; never animate layout properties.
 - **Slop animation is forbidden — motion serves meaning.** Every animation or hover must map to a real interaction, state change, or affordance. A hover that changes nothing, motion on a non-interactive element, or a decorative micro-animation with no informational purpose is slop — do not add it.
 - **Done is the `/visual-qa` dual-oracle gate, not your own glance.** A frontend design task is verified through `/visual-qa` (real browser at 375 / 768 / 1280px, every page, with interaction states and motion driven and inspected) until the dual-oracle completion gate passes on fresh evidence.
@@ -142,7 +147,7 @@ Domains: `product` `style` `typography` `color` `landing` `chart` `ux` `react` `
 | Situation | Load |
 |---|---|
 | Brand/style not among the 70 in `references/design/`, or the user says "Open Design" | `open-design` skill — the local nexu-io/open-design library (137+ design skills, 150+ design systems) |
-| Driving a browser for the Design QA phase | `agent-browser` skill |
+| Driving a browser for the Design QA phase | `visual-qa` skill: omowright from js eval (owned engine for renders, attached engine for signed-in pages) |
 | Pure TypeScript/logic work with zero visual surface | `programming` skill alone — this skill adds nothing there |
 
 ## Activation

package/packages/shared-skills/skills/frontend/references/design/README.md

@@ -7,7 +7,7 @@ You are an elite frontend design engineer. Your only job in this skill is to **r
 
 You are not done when it is clean and correct. You are done when a senior product designer at Linear, Stripe, or Supabase would ship it: surfaces read as real materials (depth, light, layering), color carries a story and a ramp, type has personality, and motion has intent. **Correct-but-flat is a failure, not a finish.** The most common way this skill fails is effort drain: the build, lint, file-size, and test gates eat all the attention and the surface ships at the floor. Spend on the surface like it is the deliverable, because for an expressive brief it is. When you render and find only bug-level issues, you are not done — you are at the start of the part that separates good from generic.
 
-Two things ship flat most often, and both read as "clean but generic": the **hero's focal object** and the **atmosphere**. Render the focal object as a real, lit, dimensional thing — a generated bitmap (imagegen) for a product/brand/object hero, or CSS/SVG art that carries light, shadow, gradient, and depth. Flat geometric primitives (plain circles and rounded rects) for a brand hero are the flat trap. Give the background depth too — gradient, glow, layered light, an atmospheric band, or a real image — not one flat fill. Glass is only one material: a dark glossy brand wants tint+blur+rim+sheen+glow, a bright playful brand wants gradient fills+soft depth shadows+a lit focal object. Pick what the brand calls for, but it must have dimension.
+Two things ship flat most often, and both read as "clean but generic": the **hero's focal object** and the **atmosphere**. Render the focal object as a real, lit, dimensional thing — a generated bitmap (imagegen) for a product/brand/object hero, or CSS/SVG art that carries light, shadow, gradient, and depth. Flat geometric primitives (plain circles and rounded rects) for a brand hero are the flat trap. Give the background depth too — gradient, glow, layered light, an atmospheric band, or a real image — not one flat fill. Glass is only one material: a dark glossy brand wants tint+blur+rim+sheen+glow, a bright playful brand wants gradient fills+soft depth shadows+a lit focal object. Pick what the brand calls for, but it must have dimension. The mechanisms for that atmosphere and for how the headline arrives come from `ambience-skill.md` (Step 10), not from memory.
 
 ## Why route at all
 
@@ -120,7 +120,7 @@ Do NOT let an expressive brief fall through to `taste-skill`. Then map the phras
 | "minimal", "clean", "Notion-like", "Linear-like", "editorial", "boring is good" | `minimalist-skill.md` |
 | "brutalist", "raw", "Swiss", "experimental", "industrial", "anti-design", "unstyled" | `brutalist-skill.md` |
 | "premium", "luxury", "calm", "expensive", "elegant", "spa", "boutique", "glossy", "glassy", "liquid glass", "startup-grade", "make it beautiful/pretty" | `soft-skill.md` + a high-craft Layer B (`supabase` / `linear.app` / `vercel` / `stripe`) |
-| "Awwwards-level", "wow factor", "magnetic", "scroll-triggered", "high-variance", "cinematic", "make it crazy" | `gpt-tasteskill.md` |
+| "Awwwards-level", "wow factor", "magnetic", "scroll-triggered", "high-variance", "cinematic", "make it crazy" | `gpt-tasteskill.md` + `ambience-skill.md` for the backgrounds and scroll scenes its sections 5 and 7 demand |
 | Neutral or operational — internal tool, dashboard, admin, "just make it usable" with no surface ambition | `taste-skill.md` as Layer A, plus the greenfield `_INDEX.md` shortlist → exactly one Layer B reference |
 
 You may also load a brand DESIGN.md from Layer B as a *concrete reference* if the user's mood maps cleanly (see the "Mood-based shortcuts" section in `_INDEX.md`).
@@ -181,6 +181,12 @@ Triggers: micro-interactions, animated components, transitions, gestures, hover/
 
 **Action:** Add `interaction-skill.md` on top of whatever style skill you selected. It anchors interaction design to the beui.dev catalog: find the nearest pattern, read its real source through the file's curl recipe, extract the mechanism (spring config, layout strategy, enter/exit order, reduced-motion path), and adapt the values to `DESIGN.md`. It owns interaction mechanics only — no visual direction — so it stacks cleanly, like `layout-skill.md`.
 
+### Step 10 — Does the work add a hero atmosphere, background, typographic reveal, or card surface effect?
+
+Triggers: an animated, shader, particle, or grid background; a hero that must feel atmospheric or dimensional; text that splits, blurs, shimmers, types, counts, scrambles, or loops; scroll-triggered chapters; spotlight / tilt / glare / glowing-border cards — or "make the hero stunning", "add a living background", "animate the headline".
+
+**Action:** Add `ambience-skill.md` on top of whatever style skill you selected. It anchors ambience to the react-bits catalog: find the nearest pattern in its intent map, read its real source through the file's curl recipe, extract the mechanism (render loop and its driver, shader uniforms or split strategy, sizing, dependencies), then run its retrofit checklist — reduced-motion static state, off-screen pause, no scroll listeners, compositor-only properties, `DESIGN.md` tokens, lazy-loaded budget. One atmosphere per hero; cursor effects stay quarantined to `gpt-tasteskill.md` briefs. It owns ambience mechanics only — no visual direction — and never vendors react-bits code.
+
 ## Stacking rules (read this once, internalize it)
 
 1. **At most one Layer A *style* skill at a time.** A layout cannot be both `minimalist-skill` and `brutalist-skill` simultaneously — they encode opposite spacing and typography philosophies. Pick one.
@@ -191,11 +197,12 @@ Triggers: micro-interactions, animated components, transitions, gestures, hover/
 6. **Layer B (brand DESIGN.md) is orthogonal to Layer A.** You can pair any Layer A skill with any Layer B brand. Use Layer B as the source of color/type/component tokens; let Layer A drive the execution discipline.
 7. **`layout-skill.md` stacks on top of any style skill** for app-shell / dashboard / split-pane work. It owns spatial structure and scroll ownership only — no visual direction — so it never conflicts with the style skill you picked.
 8. **`interaction-skill.md` stacks on top of any style skill** for interaction/motion work. It owns interaction mechanics — springs, layout morphs, enter/exit orchestration, reduced motion — and no visual direction, so it never conflicts with the style skill either.
+9. **`ambience-skill.md` stacks on top of any style skill** for hero atmosphere, backgrounds, typographic reveals, scroll chapters, and card surface effects. It owns ambience mechanics and their retrofit contract — no visual direction — so it stacks with the style skill and with `interaction-skill.md` at the same time.
 
 ## Anti-patterns — do not do these
 
 - **Don't load nothing and just freestyle.** That produces the exact "generic AI SaaS slop" — purple-blue gradient backgrounds, rounded-2xl-on-everything, three feature cards in a grid, generic Inter font, lorem ipsum. The skills exist precisely to prevent this.
-- **Don't load five files "to be safe".** That blows context and creates contradictory rules. Pick deliberately.
+- **Don't mark state with a coloured accent border.** `border-l-2 border-primary` on the selected row, a primary outline on the focused card — on a rounded surface that is the generic AI tell. State is ink-alpha washes, a glyph for selection, tonal layering for focus; `focus-visible` rings are the only coloured edge. Includes pre-existing instances on surfaces you touch.
 - **Don't ignore the user's named brand.** If they say "Linear-style" and you build something that doesn't match Linear's actual aesthetic (purple, ultra-tight spacing, mono accents, etc.), you have failed the routing.
 - **Don't apply a Layer B brand verbatim if the project is not that brand.** The DESIGN.md captures *inspiration* — extract the tokens (palette, type scale, component patterns) and apply them to the project's own content. Do not copy logos or trademarked imagery.
 - **Don't use imagegen skills to write code.** They are explicitly image-only. The agent has been observed trying to "describe" the image as React code — that is the wrong skill, switch to `image-to-code-skill.md` instead.
@@ -214,7 +221,7 @@ Once references are loaded, before writing any UI code:
 7. **Match the project's existing patterns FIRST.** If the codebase already uses CSS Modules, don't introduce Tailwind. If it uses styled-components, don't introduce CSS-in-JS variants. The references guide *style*, not *infrastructure*.
 8. **All tokens trace back to `DESIGN.md`.** No orphan hex codes, no magic px values. If you need a new token, update `DESIGN.md` first.
 9. **New reusable patterns (used 2+ times) get documented back into `DESIGN.md` Section 5.**
-10. **No generic-default drift (expressive briefs).** The shipped CSS must use the `DESIGN.md` material, not the model's priors. Load the declared fonts (do not silently fall back to Inter or system fonts), build elevated surfaces from the declared multi-layer recipe (not a lone `backdrop-filter: blur`), and color from the ramp stops (not one tint at varied opacity). Grep your styles before QA: finding `Inter`, a single blur on "glass", or one brand hex reused everywhere means you regressed to priors — fix it before declaring done.
+10. **No generic-default drift (expressive briefs).** The shipped CSS must use the `DESIGN.md` material, not the model's priors. Load the declared fonts (do not silently fall back to Inter or system fonts), build elevated surfaces from the declared multi-layer recipe (not a lone `backdrop-filter: blur`), and color from the ramp stops (not one tint at varied opacity). Grep your styles before QA: finding `Inter`, a single blur on "glass", one brand hex reused everywhere, or a coloured accent border marking selected/focused state (`border-l-2 border-primary` et al) means you regressed to priors — fix it before declaring done.
 
 ## Quick lookup table — most common requests
 
@@ -235,6 +242,7 @@ Once references are loaded, before writing any UI code:
 | "Build a dashboard / settings / inbox / app shell" | one style skill (usually `taste-skill.md`) + `layout-skill.md` |
 | "Panel won't scroll / footer pushed off-screen / mobile overflow" | Add `layout-skill.md` to current stack |
 | "Add micro-interactions / animate this / make it feel alive" | Add `interaction-skill.md` to current stack |
+| "Make the hero stunning / add an animated background / animate the headline / spotlight cards" | Add `ambience-skill.md` to current stack |
 
 ## Phase Final — Design QA (MANDATORY, runs after implementation)
 
@@ -244,6 +252,7 @@ This skill adds only the design-taste judgments `/visual-qa` cannot make for you
 
 1. **Two kinds of failure count equally — fix both, then re-check.** Defects: clipping, wrong font, missing state, jank. Flatness: a surface that reads generic next to the loaded reference. When the render is bug-free but flat, you are NOT done — RAISE the design: deepen the material layering, give the color a real perceptual ramp (multiple stops / OKLCH, not one tint at varied opacity), render the hero focal object as real dimensional material (a generated bitmap, or real light/shadow/gradient/depth — never flat geometric primitives), and add the one signature moment. Patching only bugs while the surface stays at the floor is the single most common way this skill ships clean-but-generic work.
 2. **Motion serves meaning; slop animation is forbidden.** Every interactive element must communicate its affordance and state changes — but a hover that changes nothing, motion on a non-interactive element, or a decorative micro-animation with no informational purpose is slop. Do not add it, and treat any you find as a defect. The hero may carry one signature moment; the rest of the surface earns motion only where it signals interaction or state.
+3. **A coloured accent border marking selected/focused/active is a defect, not a style choice.** Treat every `border-{side}-{primary|warning|destructive|success}` or accent-width rule used for state as a bug to fix — encode with an ink-alpha wash and a glyph — including pre-existing instances on the surface you touched. `focus-visible` rings are exempt.
 
 Report "done" only when `/visual-qa` has passed on fresh evidence AND neither a visual bug nor a floor-level or slop-laden surface remains.
 

package/packages/shared-skills/skills/frontend/references/design/_INDEX.md

@@ -59,6 +59,7 @@ Beyond the 12-file Layer A library, the design ruleset carries project-original
 | File | Purpose | Load when |
 |---|---|---|
 | `interaction-skill.md` | Interaction mechanics anchored to the beui.dev catalog: find the nearest pattern, read its real source through the curl recipe, extract the mechanism (spring config, layout strategy, enter/exit order, reduced-motion path), and adapt it to `DESIGN.md` motion tokens. | Any work adding or changing interaction or motion — micro-interactions, animated components, transitions, gestures, hover/press/state feedback, loading/success/error morphs, "make it feel alive". |
+| `ambience-skill.md` | Ambience mechanics anchored to the react-bits catalog: hero atmosphere and animated backgrounds, typographic reveals, scroll chapters, card surface effects. Find the nearest pattern in its intent map, read the source through the curl recipe, extract the mechanism, and run the retrofit checklist (reduced-motion static state, off-screen pause, compositor-only, tokens, budget). Cursor effects are quarantined to `gpt-tasteskill.md` briefs. | Any work adding a hero atmosphere, animated or shader background, text reveal (split, blur, shimmer, typewriter, count-up, marquee), scroll storytelling, or spotlight / tilt / glare / glowing-border cards. |
 | `layout-skill.md` | Layout mechanics: scroll ownership, the two silent CSS contracts, named primitives, content-stress matrix. Zero visual direction. | App shells, dashboards, split panes, or a layout that breaks under real content. |
 | `lazyweb.md` | Curl-only real-product screen research for design direction. | Greenfield design research lanes. |
 | `stylegallery.md` | Curl-only lookup of named spatial pattern contracts: primary problem, constraints, scroll ownership, anti-patterns. | A concrete spatial problem needs a documented pattern, alongside `layout-skill.md` mechanics. |