index.md

HTML wrapper for HTML-only fetchers; content is shown verbatim.

Raw file

Links (extracted)

Clickable list for reviewers; avoids copy/paste line-wrap issues.
# GGQ — Deep dive: Vercel + Anthropic “skills” (2026-01-25)

This note documents the GGQ “skills” work (Vercel Web Interface Guidelines + Anthropic/Claude frontend-design), with the exact blackboard tasks, the published artifacts, and the narrations written during that work.

---

## Executive summary (verified)

- We used **Vercel “web-design-guidelines”** as an audit checklist for static HTML packs (a11y, focus, labels, CLS, motion).
- We used **Anthropic/Claude “frontend-design”** as a design-thinking + implementation skill to produce a first‑principles prototype (then checked it against the Vercel guidelines).
- Outputs are published as hosted-static review packs (HTML-first) and include pinned skill sources + audit reports + screenshots + narrations.

---

## Public entrypoints (verified URLs)

Skills catalog (GET-only, index of skill metadata):

https://infrafabric.io/llm/skills/

Machine index:

https://infrafabric.io/llm/skills/index.json.txt

Specific skills used (catalog entries):

Vercel (skills.sh mirror) — `web-design-guidelines`:

https://infrafabric.io/llm/skills/skills.sh/vercel-labs/agent-skills/web-design-guidelines/index.json.txt

https://infrafabric.io/llm/skills/skills.sh/vercel-labs/agent-skills/web-design-guidelines/SKILL.md.txt

Anthropic (claude-code mirror) — `frontend-design`:

https://infrafabric.io/llm/skills/claude-code/anthropics/claude-code/frontend-design/index.json.txt

https://infrafabric.io/llm/skills/claude-code/anthropics/claude-code/frontend-design/SKILL.md.txt

GGQ v2 “skills A/B” pack (Vercel guidelines audit + fork + narration):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/index.html

GGQ v3-2 pack (first-principles prototype + pinned skills + audit + narration):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/index.html

---

## Blackboard tasks (actual task snapshots)

### IF-658 (Vercel Skills audit on v2 baseline)

Status: `done`  
Owner: `codex-cli (gpt-5.2) [sid: 019b4a9a-6f60-71e1-8238-20a2c1108fef]`  
Started (UTC): `2026-01-22T08:04:36Z`  
Done (UTC): `2026-01-22T08:38:39Z`

Task:

GGQ UI demos v2: Vercel Skills audit (web-design-guidelines) + punch-list report (no changes)

Acceptance:

Publish a report that records the exact skill content used (pinned), lists actionable layout/UX fixes for the v2 demo pack, and includes baseline screenshots (desktop+mobile) for reference.

Published outputs (from the task `result`):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/skill_source/SKILL.md.txt

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/skill_source/web-interface-guidelines.command.md.txt

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/report/guidelines_pin.txt

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/report/skills_audit_findings.txt

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/screenshots/AB_home_desktop.png

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/screenshots/AB_home_mobile.png

Baseline reference pack:

https://infrafabric.io/static/hosted/review/ggq/ui-demos/2026-01-22-v2/index.html

Verified note: `skills_audit_findings.txt` contains **95** findings (counted by matching lines containing `" - "`).

---

### IF-659 (skills-driven fork + narration + “after” audit)

Status: `done`  
Owner: `codex-cli (gpt-5.2) [sid: 019b4a9a-6f60-71e1-8238-20a2c1108fef]`  
Started (UTC): `2026-01-22T08:04:36Z`  
Done (UTC): `2026-01-22T08:38:55Z`

Task:

GGQ UI demos v2: skills-driven A/B fork (layout improvements) + publish separate pack

Acceptance:

Publish an A/B comparison pack: baseline links + new 'skills' fork pack with improved layout; include screenshots and a short narrative of what changed.

Published outputs (from the task `result`):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/index.html

Narration (what changed + why):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/narration.md.txt

Post-fix audit report:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/report/skills_audit_findings_after.txt

Download bundle:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/ggq-ui-demos-2026-01-22-v2-skills-pack.tar.gz

Verified note: `skills_audit_findings_after.txt` is `✓ pass`.

---

### IF-668 (v3 first‑principles pack + pinned skills + audit)

Status: `done`  
Owner: `codex-cli (gpt-5.2) [sid: 019b4a9a-6f60-71e1-8238-20a2c1108fef]`  
Started (UTC): `2026-01-22T14:58:38Z`  
Done (UTC): `2026-01-22T15:13:05Z`

Task:

GGQ UI demos v3: first-principles redesign prototype (competitive brief) + skills audit

Acceptance:

Publish a new hosted-static GGQ UI demos v3 pack (separate from v2/CT250) guided by ui-competitive-landscape + ui-layout-proposals, with: (1) search-first home hero (autosuggest + keyboard accessible), (2) 3-level Destinations overlay in 3 columns (region/subregion/place), (3) redesigned Destination + Article pages with print-like typography, (4) Map split-view page mock, (5) ads treated as designed components (leaderboard + right-rail; no overflow; labeled), (6) Vercel web-interface-guidelines audit report for v3 with 0 findings or documented exceptions, (7) desktop+mobile screenshots, and curl-verifiable URLs.

Published outputs (from the task `result`):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/index.html

Markdown overview:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/README.md.txt

Narration:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/narration.md.txt

Pinned skill sources (v3-2):

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/skill_source/claude-frontend-design.SKILL.md.txt

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/skill_source/vercel-web-interface-guidelines.command.md.txt

Pinned hashes:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/skill_source/claude-frontend-design.SKILL.md.sha256

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/skill_source/vercel-web-interface-guidelines.command.md.sha256

Audit report:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/report/skills_audit_findings.txt

Screenshots:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/screenshots/home-desktop.png

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/screenshots/home-mobile.png

---

### IF-653 / IF-655 (“if.skills” meta)

These were the meta tasks created to formalize an `if.skills` pack and link it into `/llm`:

- IF-653 — `in_progress`
- IF-655 — `todo`

Verified note: the live catalog that exists today is `/llm/skills/` (above). The path referenced in the IF-653 title (`/llm/products/if-skills/`) is currently not present as a published endpoint.

---

## What “skills” meant in this GGQ work (black/white)

### Verified (what exists)

- A pinned copy of the Vercel guidelines command (the rules) was shipped inside the GGQ packs (see the `skill_source/*.command.md.txt` URLs above).
- Audit outputs were saved as plain files in each pack (`report/skills_audit_findings*.txt`) and can be counted/checked without any tooling.
- Narrations exist as published Markdown text files (`narration.md.txt`) and are readable without downloads.

### Interpretation (how we used it)

- We treated Vercel’s guidelines as a **non-negotiable quality gate** for UI mechanics (labels, focus, image dimensions, reduced motion).
- We treated the Anthropic “frontend-design” skill as a **design-direction tool** (pick an aesthetic + execute it), then validated the result against the Vercel checklist to avoid regressions.

---

## Narration: which session wrote it + where it is

### Session (verified via blackboard task owners/session_ids)

`019b4a9a-6f60-71e1-8238-20a2c1108fef`

This session id appears on IF-658 / IF-659 / IF-668, and these tasks are the ones that produced the narrations.

### Published narrations (verified URLs)

GGQ v2 skills A/B narration:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-skills/2026-01-22-v2-skills/narration.md.txt

GGQ v3-2 narration:

https://infrafabric.io/static/hosted/review/ggq/ui-demos-v3/2026-01-22-v3-2/narration.md.txt

---

## Runbook: how future agents find + reuse the skills

1) Start from the canonical GET-only index:

https://infrafabric.io/llm/skills/

2) Use the machine index to search for skill ids:

https://infrafabric.io/llm/skills/index.json.txt

3) For GGQ work, prefer *pinned, pack-local* copies (to avoid upstream drift):

- V2 skills pack `skill_source/` (Vercel audit inputs)
- V3-2 pack `skill_source/` (Anthropic design skill + Vercel audit inputs)

4) Publish-safe practice (repeatable):

- Pin the skill input file(s) you used in `skill_source/` and include sha256 (either `.sha256` files or a small pin note like `guidelines_pin.txt`).
- Record the findings output as a plain file under `report/` (so it’s diffable and countable).
- Add a narration that explains what changed + why (so future agents can reproduce intent, not just outcomes).