gogh

Gogh

Source cited Obsidian operating brain that turns six frontend design skills into one advisable stack.

Watch

Watch here

Gogh is an evidence-gated Obsidian brain for AI coding agents that need better frontend design judgment than a single prompt can provide. It combines six curated skills into stack advice: which skill to use when, how to combine them, how to resolve conflicts, and how to keep a unified anti-slop rulebook tied to dated sources.

Current maturity: market-ready pending the release gate. The structural audit passes at market-ready; the final release is cut by the operator procedure.

Promise

Turn six frontend design skills into one source-cited operating brain for stacking taste prompting, micro-interaction execution, toolchain enforcement, retrieval, and audit work.

Buyer

Developers, indie hackers, design engineers, and agencies who build frontends with AI coding agents (Claude Code, Cursor, Codex, Windsurf) and want the output to look intentional and brand-fit rather than templated AI slop.

Six Skills

Skill	Author	Mechanism	License
taste-skill v2, design-taste-frontend	Leon Lin	Taste prompting for macro aesthetic direction, three dials at 8/6/4, and Section 14 pre-flight	MIT
make-interfaces-feel-better	Jakub Krehel	Micro-interaction execution for press states, concentric radii, and motion values	MIT (README-declared)
Impeccable v3.9.1	Paul Bakaus	Toolchain enforcement through commands, PRODUCT.md and DESIGN.md contracts, a deterministic 45-rule detector, and edit-time hooks	Apache-2.0
Anthropic frontend-design	Anthropic	Taste prompting baseline with committed aesthetic direction	Apache-2.0 (per-skill LICENSE.txt)
ui-ux-pro-max v2.10.2	nextlevelbuilder	Retrieval database for styles, palettes, font pairings, and UX rules	MIT
Vercel web-design-guidelines	Vercel Labs	Audit layer with runtime-fetched rules and file:line findings	MIT (ruleset repo; skill wrapper README-declared)

Mechanism map: taste prompting from taste-skill and Anthropic frontend-design, micro-interaction execution from make-interfaces-feel-better, toolchain enforcement from Impeccable, retrieval from ui-ux-pro-max, and audit from Vercel web-design-guidelines.

Showcase

Six message-match variants of one landing page, each a distinct editorial format instead of a generic template. Full source in examples/aimh-pro-landing-variants/, live at agricidaniel.github.io/gogh.

Outputs

• assets/template-brain/, the distributable Obsidian template vault
• SKILL.md plus scripts, the agent-facing operating layer
• Stack advisor reports that translate a project brief into recommended skills, order of use, dial settings, conflicts, and audit checks
• Cheat sheet, stack decision one-pager, unified pre-flight mega-checklist, and quickstart
• Registry diff reports for upstream captures when the adapter wave lands

Quick Start

python -m pip install -e .
gogh demo
gogh lint --vault examples/sample-vault
gogh report --vault examples/sample-vault --html-only

To create a client vault:

gogh new acme --client-name "Acme Co" --owner "Daniel Agrici" --out-dir ~/gogh-vaults
gogh ingest --vault ~/gogh-vaults/acme --file tests/fixtures/sample-source.md
gogh synthesize --vault ~/gogh-vaults/acme
gogh visuals --vault ~/gogh-vaults/acme
gogh report --vault ~/gogh-vaults/acme --html-only
gogh next --vault ~/gogh-vaults/acme

Boundaries