landing-page-replication-v5
Skill0 starsChanged 26 days agoMITView on GitHub
Use when replicating marketing/SaaS/D2C landings or WebGL/Lenis/scroll-theater portfolios to measurable fidelity. v5 keeps v4's Loop 5 Behavior gates and closes three "missing consequence" relapses: replica-only-motion blocker, continuous pointer-theater probe (no :hover false positives), broad-spectrum scroll-state sampling. Prefer --behavior-offline. Capture → signal → skeleton → density → micro-parity → behavior → polish.
- Installs packages
In this repository15 other files it gives its agents.
- ai-native-mihomoSkill
- ai-native-proof-ledgerSkill
- clerk-authSkill
- cloudflare-cost-guardSkill
- design-md-visual-systemSkill
- doubao-ttsSkill
- fleet-orchSkill
- geo-job-huntSkill
- github-telegram-feedback-pipelineSkill
- interactive-projects-streamSkill
- seo-masterSkill
- talkcutSkill
- typesafe-aiSkill
- video-script-conversionSkill
- webgl-threejs-background-animationSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 45
Do not invent replica-only shaders when a pack adapter already covers the mechanism.
- Distinctive: line 46
Do not skip Loop 5 because the pack recipe “looks right.”
- Very specific: line 55
If unsure → Ship. Replicate design *language*.
- Distinctive: line 75
Two-Reference Rule: before Loop 1: positive target + anti-reference + 3 sharp edges.
- Very specific: line 97
Write docs/SIGNAL.md from references/signal-sheet.md.
- Distinctive: line 121
Do not invent replica-only motion unless listed as a sharp edge — now a
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Landing Page Replication (v5)
Domain-agnostic 7-phase pipeline (Loop 0–6). Case specifics live in
references/case-*.md. Load references on demand.
Why v5: v4 taught "missing consequence, not missing information" — then
relapsed three levels deeper (docs naming a blocker with no function; pointer
probe beaten by ordinary :hover; scroll-state hardcoding haoqi fields;
self-test as string theatre; --behavior bombarding the target). v5 makes
those gates real. Detail: references/v5-build-log.md.
Worked example (haoqi): Hero IMR Δ 3% (pass) + SLR 0.52 → FAIL
(--scroll-length). Fix tunnel height before ±2px chase. Gallery J.
0. Sibling assets (design-templates)
This skill is the measurement pipeline. Visual leftovers live in
lizliz404/design-templates
(local: D:\dev\repos\design-templates). Map: templates/sibling-routes.md.
Need after capture tagged WEBGL_THEATER / agent chrome | Load |
| Paper / mesh atmosphere (not a status orb) | templates/paper-shaders/SKILL.md |
| One thinking / glass orb | templates/ui-patterns/thinking-orb/SKILL.md (MIT wgsl+metal already grabbed) |
| Agent thinking / stream / approval / dense tables | templates/design/beautiful-ui-ai-interfaces/ |
| Full-bleed Three.js that blends into the page | sibling skill webgl-threejs-background-animation in this repo |
| One Community 3D product hero | templates/ui-patterns/threeui-hero-adapter.md — not a diagram |
Do not invent replica-only shaders when a pack adapter already covers the mechanism. Do not skip Loop 5 because the pack recipe “looks right.”
1. Risk posture + legal
| Posture | When | Behavior |
| Study / internal-only | Never ships publicly | Single positive target OK; pick one anti-reference |
| Ship as your product | Live commercial product | Composite 3–5 leaders; never name source publicly; then brand-distinction-v1 |
If unsure → Ship. Replicate design language; never copy logos, proprietary
imagery, or hotlink target fonts. (Not legal advice.) Full IP notes:
references/ip-and-fonts.md.
2. Pipeline
Capture → Signal Sheet → Skeleton → Density → Micro-parity → Behavior → Polish
│ │ │ │ │ │
capture.py SIGNAL.md tokens --imr sharp edges --behavior-offline ← default
capture-runtime.py --scroll-length (--behavior live = supplemental)Gates: pipeline.yaml. Checklists: references/fidelity-checklist.md.
Behavior matrix: references/behavior-gates.md.
Loop 0 — Capture
Reconstruct story arc, type/color roles, density, and runtime surface.
Two-Reference Rule before Loop 1: positive target + anti-reference + 3 sharp edges.
python3 scripts/capture.py --url https://target.com --out recon/
python3 scripts/capture-runtime.py --url https://target.com --out recon/runtime.json
python3 scripts/validate-capture.py recon/Route by fingerprint (references/capture-router.md). Bot Management is the
2026 default failure mode — do not re-hit the live target later for verification.
Gate: validate-capture.py PASS; ≥5 screenshots; runtime.json with
canvasCount / webgl / maxScroll / hasLenis (or static N/A). Tag
WEBGL_THEATER when canvasCount > 0.
Loop 1 — Signal Sheet
python3 scripts/extract-tokens.py recon/index.html --base-url https://target.com
python3 scripts/validate-signal.py docs/SIGNAL.md recon/runtime.jsonWrite docs/SIGNAL.md from references/signal-sheet.md. Must include
Interaction Contract table (evidence / strategy / pass criterion) for every
P0 runtime signal — not prose. Template + fields: signal-sheet.md.
Loop 2 — Skeleton
Tokens for roles used 3+ times. Reserve scroll-tunnel height from
offsets.json / runtime.json.maxScroll. Gate: audit.py --typescale.
Loop 3 — Density
Pin shots at 0 / 25 / 50 / 75 / 100% scroll. Cut a weak section before adding more.
python3 scripts/audit.py --imr target.png replica.png --tolerance 8
python3 scripts/audit.py --density {replica_url}
python3 scripts/audit.py --scroll-length recon/offsets.json {replica_url} --min-ratio 0.85Scroll-length fail blocks exit unless docs/GAPS.md waiver (Gallery J).
Loop 4 — Micro-parity
Implement Two-Reference sharp edges. Font metrics: metric-align-fonts.py.
Do not invent replica-only motion unless listed as a sharp edge — now a
machine blocker in Loop 5 (replica-only-motion).
Loop 5 — Behavior
Prove every P0 interaction live — not via frozen PNG.
# Recommended default (1 replica session; target from runtime.json):
python3 scripts/audit.py --behavior-offline recon/runtime.json {replica_url} \
--signal docs/SIGNAL.md --out recon/behavior-diff.json
python3 scripts/audit.py --reduced-motion {replica_url}
# Live re-probe of target only when runtime.json is stale (1 shared target session):
# python3 scripts/audit.py --behavior {target} {replica_url} --signal docs/SIGNAL.mdBlockers (all implemented in audit.py — see references/behavior-gates.md):
canvas | webgl | scroll-length | pointer-theater | scroll-states |
reduced-motion | smooth-nav | replica-only-motion
Theater-runtime-first priority: scroll length → theater pointer/scroll → sticky morph → nav/tabs → reduced-motion → a11y.
Gate: behavior-diff.json blockers empty or each deferred in GAPS with
pass criterion. Gallery I fails hard.
Loop 6 — Polish
a11y → LCP → ±2px on P0 only. Do not polish while Loop 5 blockers remain.
3. Tool routing
| Need | Tool |
| Capture / runtime | capture.py, capture-runtime.py, validate-capture.py |
| Tokens / signal | extract-tokens.py, validate-signal.py |
| Density / type | audit.py --imr/--diff/--typescale/--density/--rhythm |
| Scroll-length | audit.py --scroll-length |
| Behavior (default) | audit.py --behavior-offline |
| Behavior (live supplemental) | audit.py --behavior (merged sessions) |
| Motion guards | --reduced-motion / --smooth-nav / --replica-only |
| Fonts | metric-align-fonts.py |
| Critique prompts | prompts/density-critique.md, prompts/behavior-audit.md |
| Self-test | run-evals.py --self-test (executes subcommands on fixtures) |
4. Common failure modes
| Failure | Fix |
| 403 / bot challenge | Manual DevTools — legitimate Loop 0 |
| IMR pass, live page dead | Loop 5; Gallery I |
| Page ~40% shorter | --scroll-length; Gallery J |
| Inner scroller missed | Gallery K |
| Long page, one scroll state | Gallery L (broad-spectrum sampling in v5) |
Ordinary :hover "proves" theater | v5 continuous theater-node probe |
| Replica-only cursor trail | --replica-only / Loop 5 blocker |
--behavior hammering target | Use --behavior-offline |
| Self-test string theatre | v5 --self-test executes audit subcommands |
5. Output layout
recon/ index.html headers.txt css/ screenshots/ computed.json dom.html
runtime.json assets.json offsets.json behavior-diff.json
docs/ SIGNAL.md GAPS.md
src/ tokens + primitives + sections + animations/ + data.ts6. References (load on demand)
| File | When |
behavior-gates.md | Loop 5 blocker matrix + probe design |
taste-rubric.md / failure-gallery.md | Density + runtime syndromes |
capture-router.md | Non-SSR / bot-blocked Loop 0 |
gaps-template.md / signal-sheet.md / fidelity-checklist.md | Contracts + exits |
ip-and-fonts.md / product-theater.md | Fonts, trade dress, theater stand-ins |
case-*.md / domain forks | Evidence-tiered case notes |
v5-build-log.md | What changed vs v4 |
7. Verification
pip install -r scripts/requirements.txt
python3 scripts/validate-capture.py recon/
python3 scripts/validate-signal.py docs/SIGNAL.md recon/runtime.json
python3 scripts/audit.py --imr target.png replica.png --tolerance 8
python3 scripts/audit.py --scroll-length recon/offsets.json http://127.0.0.1:5173 --min-ratio 0.85
python3 scripts/audit.py --behavior-offline recon/runtime.json http://127.0.0.1:5173 \
--signal docs/SIGNAL.md --out recon/behavior-diff.json
python3 scripts/run-evals.py --self-test8. Quality checklist
- Loop 0:
validate-capture.pyPASS; runtime.json or static N/A - Loop 1:
validate-signal.pyPASS; Interaction Contract covers P0 flags - Loop 3: IMR mid-scroll bands; SLR ≥0.85 or GAPS row
- Loop 5:
--behavior-offlineblockers empty or each in GAPS - Loop 5:
--reduced-motionPASS; smooth-nav when target smooth-scrolls - No undeclared replica-only motion
run-evals.py --self-testPASS (subcommand execution, not grep-only)- P0 gaps closed; P2/P3 listed in GAPS.md
9. v4 → v5 delta
| Hole in v4 | v5 fix |
replica-only-motion named in 4 docs, absent in _parity_report | _replica_only_motion + blocker |
Pointer 2-point / hero-button :hover false positive | ≥3-point continuous delta on canvas/[data-theater]/[data-pointer] only |
| Scroll-state sampling hardcoded to haoqi fields | Broad transform/opacity/filter sampling; body class/bg fallback only |
--self-test string-matching theatre | Executes each audit subcommand on fixtures |
--behavior opens 6–7 Chromium sessions against target | Target probes merged to 1 session; --behavior-offline is default |
| case-*.md evidence weight unclear | Evidence tier labels |
| Fake logo-cloud brand collision | Note in ip-and-fonts.md |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

