reimagine-it
Kayforkind/reimagine-it/docs/llms.txt
Content-Derived Design: an engine that reads an HTML file and redesigns it from its own content — palette, motif, and motion derived from the nouns, dates, numbers, emails, and hex colors already in the source. Source facts are never invented; only the composition changes. Ships one offline HTML file. The source file is the brief. A bakery cannot come out marine-teal. Seventeen design directions (tokens) exist: webpage, infographic, svg, 3js, simulation, artistic, cinematic, dashboard, photography, landing, editorial, gradient, showcase, motion,…
llms.txt190 starsChanged 26 days ago
# reimagine-it
> Content-Derived Design: an engine that reads an HTML file and redesigns it from its own content — palette, motif, and motion derived from the nouns, dates, numbers, emails, and hex colors already in the source. Source facts are never invented; only the composition changes. Ships one offline HTML file.
The source file is the brief. A bakery cannot come out marine-teal. Seventeen design directions (tokens) exist: webpage, infographic, svg, 3js, simulation, artistic, cinematic, dashboard, photography, landing, editorial, gradient, showcase, motion, glass, lookbook, particles. **Auto** scores every lane against the content and picks the strongest.
Not Keith Mangold’s reimagineit.ai interview SaaS. This is Kayforkind/reimagine-it on npm and GitHub.
## Install
```bash
npx skills add Kayforkind/reimagine-it # agent skill
npx reimagine-it --auto -i page.html -o redesign.html # plain CLI, no agent needed
```
## Commands
| Command | What it does |
|---------|--------------|
| `npx reimagine-it --auto -i page.html -o out.html` | Choose, generate, verify, explain |
| `npx reimagine-it -i page.html -t infographic -o out.html` | One explicit direction |
| `npx reimagine-it extract -i page.html -o signals.json` | The facts the engine reads, as JSON (no redesign) |
| `npx reimagine-it variations -i page.html -n 4 -o review/` | Several directions + a contrast sheet |
| `npx reimagine-it lock -i brand.html -o house.lock.json` | Capture a design surface as reusable data |
| `npx reimagine-it -i page.html --ref house.lock.json -o out.html` | Redesign inside an existing brand |
| `npx reimagine-it audit out.html` | 19 deterministic craft checks (exit 0/1/2) |
| `npx reimagine-it mcp` | MCP server, 8 tools, stdio |
## Rules an agent must follow
1. Run Auto first (`--auto`); read the printed rationale.
2. Never invent facts — every date, number, email, and link in the output exists in the source. `extract` is the audit surface for this.
3. Fidelity must be ≥ 80% (the engine prints the percentage of detected source facts preserved; all committed examples hold 100%).
4. `audit` must report zero failures before the result is called done.
5. Output is a single offline HTML file — no CDN, no external fonts unless `--web-fonts`.
6. `--output` may never be the source file (the CLI exits 2); pass `--no-clobber` when an existing output must not be replaced. Auto runs on one output path are serialized by an advisory lock; `--force` overrides it.
## For MCP hosts
```json
{ "mcpServers": { "reimagine-it": { "command": "npx", "args": ["reimagine-it-mcp"] } } }
```
Tools: `reimagine`, `design_auto`, `design_variations`, `design_lock`, `extract_content`, `list_tokens`, `audit_html`, `list_rules`. Per-host setup: https://github.com/Kayforkind/reimagine-it/blob/main/docs/mcp-hosts.md
## Key links
- Live playground (runs in the browser, no install): https://kayforkind.github.io/reimagine-it/#playground
- Nine committed journeys with auditable reports: https://github.com/Kayforkind/reimagine-it/tree/main/examples/end-users
- Public-domain proof lane (NPS, NASA, NOAA, Census, Federal Register, Smithsonian pages rebuilt by Auto): https://github.com/Kayforkind/reimagine-it/tree/main/examples/public-sources
- Case studies: https://github.com/Kayforkind/reimagine-it/blob/main/docs/CASE_STUDIES.md
- Method (Content-Derived Design): https://github.com/Kayforkind/reimagine-it/blob/main/docs/MANIFESTO.md
- Contributing (community proof cases welcome): https://github.com/Kayforkind/reimagine-it/blob/main/CONTRIBUTING.md
## Optional
- Machine-readable skill metadata: [AGENTS.md](https://raw.githubusercontent.com/Kayforkind/reimagine-it/main/AGENTS.md) and `skills/reimagine-it/SKILL.md` frontmatter (`trigger_phrases`, `capabilities`, `keywords`).
- Dataset of {source, signals, output} tuples for design-model research: `scripts/build-dataset.js` in the repo.
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
Posts are public.Sign in to post
No one has posted yet. Be the first.

