Agent skill

Reimagine It

by Kayforkind in Kayforkind/reimagine-it

Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file.

MITAuto-check passedDocuments & Office

Install Reimagine It

skills CLI
$ npx skills add Kayforkind/reimagine-it --skill reimagine-it -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Kayforkind/reimagine-it reimagine-it --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Kayforkind/reimagine-it.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/reimagine-it .claude/skills/reimagine-it && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
reimagine-it
GitHub stars
205
Token cost
~11k tokens
SKILL.md length
5,250 words
Files
38 (incl. references)
Skills in repo
7
Repo updated
First seen
Licence
MIT

At a glance

Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file.

  • Works in 8 steps: Token parse (open brief — do not drop… → Context sniff (do not skip) → Four notes (private) → …
  • The user says /reimagine-it
  • SKILL.md covers Engine — run the CLI; do not…, Categories (you choose · agent…, Hard contract and Procedure, plus 1 more section
  • Calls npx, git and just

What it does

Reimagine It is an agent skill from Kayforkind/reimagine-it. Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file. CLI: npx reimagine-it --auto -i page.html -o out.html; also variations, lock, and audit. Use when the user says /reimagine-it, "reimagine it", "reinvent this", "redesign this page", "make an infographic", wants a content-aware redesign or a visual leap instead of a mood board. All 17 tokens: webpage, infographic, svg, 3js, simulation…

Its SKILL.md is about 11k tokens, which your agent loads only when the skill is triggered. The skill folder holds 40 other files, including reference files (for example `examples.md`, `references/craft-floor.md` and `references/domains/README.md`).

It sits in Documents & Office, covering Infographics, Word documents and PowerPoint presentations. It works with Microsoft PowerPoint and Microsoft Word. The repository describes itself as: Kayforkind Content-Derived Design CLI. HTML in, standalone HTML out from the source's own nouns, dates, numbers, and colors. Not the reimagineit.ai interview SaaS. The licence is MIT.

When your agent uses it

  • The user says /reimagine-it
  • Redesign this page
  • Make an infographic
  • Wants a content-aware redesign

Example prompts

  • “reimagine it”
  • “reinvent this”
  • “redesign this page”
  • “/reimagine-it”

Requirements

  • Python 3
  • Node.js

Workflow steps

8 steps, taken from the step headings in SKILL.md.

  1. Token parse (open brief — do not drop words)
  2. Context sniff (do not skip)
  3. Four notes (private)
  4. Form router
  5. Build
  6. full plus-pass
  7. Verify (functional + visual)
  8. Report

What it can do on your machine

Read from SKILL.md and the folder at commit e8c9378. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • npx
    • git
    • just
    • python

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use npx and git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Reimagine It loads about 11k tokens when it runs, and up to ~57k if it reads all its reference files. Until then it costs about 232 tokens; SKILL.md has 5,250 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~232
When it runs · the whole SKILL.md, loaded when a task matches
~11k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~57k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Kayforkind/reimagine-it at commit e8c9378, republished under its MIT licence (© Kayforkind). 5,250 words, ~10,892 tokens.

Download SKILL.mdSave it as .claude/skills/reimagine-it/SKILL.md (or your agent's skills folder). This skill also uses 37 other files; get the full folder from GitHub.
name
reimagine-it
description
Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file. CLI: npx reimagine-it --auto -i page.html -o out.html; also variations, lock, and audit. Use when the user says /reimagine-it, "reimagine it", "reinvent this", "redesign this page", "make an infographic", wants a content-aware redesign or a visual leap instead of a mood board. All 17 tokens: webpage, infographic, svg, 3js, simulation, artistic, cinematic, dashboard, photography, landing, editorial, gradient, showcase, motion, glass, lookbook, particles — plus leftover words as a brief. Also /reimagine-it audit for quality checks. The agent runs the CLI; it does not hand-author the page. Installed via npx skills add Kayforkind/reimagine-it or plugin marketplace. Not /better. Not a mood board. Does not ingest PDF, PPTX, or DOCX.
license
MIT
metadata.author
Kayforkind
metadata.version
2.15.0
metadata.hosts
claude-code, cursor, codex, copilot, gemini-cli, factory-droid, agent-skills
keywords
content-derived-design, redesign, infographic, svg, threejs, frontend, html, css, palette, motion, offline
category
Design
capabilities
content-extraction, color-palette-generation, infographic-creation, svg-generation, threejs-scene-generation, simulation-generation, accessibility-audit…
trigger_phrases
/reimagine-it, reimagine it, reinvent this, redesign this page, make an infographic, content-aware redesign, design from content, redesign from source…

/reimagine-it

Banks: references/notes.md · references/forms.md · references/webpage-craft.md (only for the webpage form) · references/craft-floor.md (the interaction contract every webpage output must clear; read every run) · references/review.md (§5.d named-object accuracy + clone scan; load on every verify) · references/forms/ (non-webpage form packs: pdf, document, slides, universal) · references/domains/ (only when the user gave a domain token: artistic / dashboard / photography / cinematic (aka 3d, webgl) / ecommerce / landing / portfolio / infographic) · references/modifiers/ (only when the user gave a modifier: glassmorphism / bento / neon / brutalism / neumorphism / handdrawn) · references/locks/ (loaded on --ref <name>) · references/research/web-craft-2025.md (deep source pack — Awwwards SOTY stack scan, Rauno/Emil craft floor, Lupi/Fragapane data humanism, Feixen/Weingart/Troxler print grammar, Apple AIDA cinematic, view-transitions, scroll-driven animations, kinetic type, sound, neubrutalism — read once, cite in reports) · references/research/infographic-craft.md (15-source infographic pack — Cleveland–McGill, Tufte, Bertin, ISOTYPE, Minard/Snow, Cairo, Lupi, FT Visual Vocabulary, USWDS, WCAG charts, InfoAlign layouts — load when the form or domain is infographic) · examples.md

/reimagine-it is a Content-Derived Design engine. The shipped product reads HTML and writes HTML: extract the headings, facts, names, dates, numbers, links, emails, and colors already in the file, then generate a standalone page in one of 17 tokens (or --auto). Default output is offline.

Do not claim the CLI redesigns PDF, PPTX, DOCX, MOBI, or “any file.” If the user pointed at another format, convert or wrap it as HTML first, run the engine, and only then — if they asked and the toolchain exists — export with a host tool. Report that extra step as host conversion, not as an engine feature.

If local.md exists beside this file, Read it after this skill (host chairs, org, paths). Public installs have no local.md.

Say once: "Running /reimagine-it."

Engine — run the CLI; do not invent the page

The shipped product is the CLI. /reimagine-it is orchestration around it, not a license to hand-author HTML.

  1. If the source is not HTML, wrap or extract it into HTML first. Do not claim the engine ingested PDF, PPTX, or DOCX.
  2. Run the matching command and use the file it writes:
bash
npx reimagine-it --auto -i page.html -o redesign.html
npx reimagine-it -i page.html -t infographic -o poster.html
npx reimagine-it variations -i page.html -n 3 -o review/
npx reimagine-it extract -i page.html
npx reimagine-it audit redesign.html

The CLI enforces the source guard mechanically: an --output path that resolves to the source file exits 2 without writing. Pass --no-clobber to also refuse replacing an existing output file — useful when an agent or CI loop must not clobber a reviewed artifact. Auto mode also holds an advisory run lock next to the artifact, so a second agent targeting the same output fails fast instead of interleaving writes; crashed runs self-heal via stale-lock detection, and --force overrides.

  1. Open that artifact. Report from it. If npx reimagine-it cannot run, ship REIMAGINED: partial and name the exact blocker — do not substitute a model-written page.

Categories (you choose · agent decides)

Optional tokens. Combine freely. You pick tokens; the agent picks questions, form (if unset), mutations, and the stretch.

CategoryYou chooseAgent decides
(none)Default. No interview.Infer, lock, form, build.
Auto — auto / --autoLet the tool manage the normal redesign loop.Extract evidence → rank coherent forms → generate candidates → verify → return the strongest artifact. Source stays read-only.
interviewTalk before build.Which questions, recommended answers, when to stop.
Form family — code cli protocol demo prose product architecture experimentForce that form family.How the notes land in it.
Visual form — svg 3js infographic canvas html webpage simulationForce a visual form.Craft inside that medium. Token infographic also loads references/domains/infographic.md (poster of an argument, not a dashboard). Token svg / 3js ship alive-micro by default (2–4 fact-tied loops that beautify; brief still freezes). Token simulation ships a playable model of facts already in this source — default paused on the first fact; nested short spans are inspectable; type in the gutter. None of these are a Texas skin; leftover words are a lens.
Host conversion (not CLI tokens) — pdf document slidesOnly if the user asked and local tools exist.Convert/wrap as HTML → run the engine → optional export (Weasyprint, python-docx, python-pptx). The CLI does not read or write these formats.
Domain aesthetic — second word after webpage: artistic dashboard photography cinematic (3d, webgl) ecommerce landing portfolio infographicForce a webpage aesthetic.Load references/domains/<domain>.md and extend the webpage-craft spine. cinematic upgrades the 3D floor to inline WebGL2. infographic is a statistical poster (common-scale encodings + ISOTYPE + data table) — not an ops dashboard.
Modifier — third word or --style <name>: glassmorphism bento neon brutalism neumorphism handdrawnLayer a UI/UX modifier on top of the domain.Load references/modifiers/<name>.md; modifiers waive matching cut-list entries and add their own non-negotiables.
Font override — --font "Family, Fallback, generic"Pin the display / body font family.Build a full font stack; degrade gracefully when the family is not on the reader's box. Never fetch a webfont at run time unless you also pass --allow-fetch.
Sound — --sound [tier] (ambient / feedback / full, default feedback)Turn on sound design for this run. Off by default.Load Howler.js sprite pack; enforce earcon tier hierarchy (§ craft-floor #7); ship visible mute + volume controls; never autoplay; provide visual alternative for every cue. Downgrades to partial if any of those are missing.
Lock / reuse — lock <path> [as <name>] · --ref <name> · --list-refsCapture a shipped output as a reusable reference, or apply one.Extract palette + type stack + motifs + motion + 3D signatures into references/locks/<name>.md; on --ref, load that pack as if it were a domain.
--notesShow the four notes.Which four.
--plan-onlyNo files.Lock + notes + form + stretch.
--fullExtra plus-pass after the hero.One mutation that serves the idea, then ship.
--variants NAsk for N distinct outputs from the same brief instead of one.Pick N distinct make-strange moves; write to .../var-1/, .../var-2/, ...
--seed <n>Pin the creative variation sample so two runs produce the same draw. Default is fresh every run.Use the seed to deterministically pick one option per variation axis.
--variant a|b|c|...Shorthand for a named seed (a = first canonical draw, b = second, …).Reproduce a specific shipped draw; useful for locks (--variant b --lock).
<brief>Any leftover words after known tokens. Open vocabulary — not a theme catalog.Treat as a creative lens on register, ground, motif, pattern, type, motion. Still sniff context; brief does not replace source facts.
Audit — auditRun deterministic quality checks on an HTML file.Runs npx reimagine-it audit — 19 checks (src/audit.js) across typography, palette, motion, content, structure, and performance with no LLM and no API key. Python scripts/audit.py is the CI mirror. Zero failures must pass. Supports --verbose and --json.

Combine freely. Known form / domain / modifier tokens load packs. Every other word is kept and followed. There is no list of allowed leftover words.

/reimagine-it auto
/reimagine-it webpage cinematic

/reimagine-it webpage infographic
/reimagine-it infographic
/reimagine-it infographic <any leftover words>
/reimagine-it svg
/reimagine-it audit
/reimagine-it 3js
/reimagine-it simulation
/reimagine-it webpage artistic glassmorphism --font "Playfair Display, serif"
/reimagine-it lock gold/domains/cinematic/after.html as house-cinema
/reimagine-it webpage --ref house-cinema
/reimagine-it webpage bento --variants 3

Interview is off unless they picked interview. Do not grill unprompted. Do not ask "what style do you want?" — not even in interview.

Every mode must ship an artifact (unless --plan-only). A list of vibes is not the deliverable.

Hard contract

  1. Read this file. Load references/notes.md and references/forms.md.
  2. Sniff context. Interview only if interview was chosen.
  3. Name the adjacent possible: spare parts already here + one combination they did not request.
  4. Pick four notes in private (device · leap · craft · effect). Answer with mutations, not mood words.
  5. Route a hero form. Run the CLI into that form (-t or --auto). Do not hand-author a replacement page.
  6. Name one stretch they did not know was in bounds. Build it if cheap; otherwise give the exact next command.
  7. Kill list in notes.md. No "wow". No prompt-slop. No TED-over-B-roll.
  8. Paid gate: code, SVG, HTML, PDF, docx, pptx, local demos are free. Ask before billed image/video/model APIs.
  9. No commit/push unless asked.
  10. Report REIMAGINED: shipped | partial | blocked.

Procedure

REIMAGINED Progress:
- [ ] 0. Mode + categories parsed (leftover words kept as brief) + context sniffed
- [ ] 0.5. Interview only if that category was chosen
- [ ] 0.75. Adjacent possible named (private)
- [ ] 0.85. Anchor list — extract 3–5 concrete nouns/proper nouns/dates/verbs from the source; every plate must map back
- [ ] 1. Four notes chosen (private)
- [ ] 2. Hero form routed (unless user forced one)
- [ ] 2.4. Variation sample picked (avoid previous draw; pin if --seed / --variant)
- [ ] 2.5. Modifiers / font override / --ref loaded if present
- [ ] 2.6. Output is standalone HTML (host conversion only if the user asked and the toolchain exists)
- [ ] 2.7. Craft floor loaded (references/craft-floor.md) — every rule enforceable before render
- [ ] 3. Hero artifact written (or N variants if --variants)
- [ ] 4. Stretch named (and built if cheap)
- [ ] 4.5. --full plus-pass if requested
- [ ] 5. Verify with evidence (functional + visual + craft-floor scan)
- [ ] 6. Report REIMAGINED: shipped | partial | blocked
Auto mode — the automatic design loop

auto is the person-like invocation: the user gives context, not a pile of commands. Infer the best form from the source, generate up to three coherent candidate directions, run the craft-floor checks, and return the strongest standalone artifact plus the selected token, seed, evidence, and rejected candidates. Never overwrite the source. Never invent facts. If a host can render previews, show the selected artifact and make the alternatives available for comparison.

Auto is deliberately model-agnostic. An agent host may use the returned plan as structured context, delegate visual review to a subagent, or supply a model-specific renderer. The core engine remains deterministic and safe without an API key. --auto is an explicit CLI equivalent; plain /reimagine-it remains the full inferred workflow.

0. Token parse (open brief — do not drop words)

This skill is a creative engine. The user may type anything after /reimagine-it. Do not require a matching pack. Do not autocorrect unknown words into a listed domain. Do not ask "what style do you want?"

  1. Split on whitespace. Consume flags and their arguments (--seed, --style, --font, --ref, --variants, --sound, --notes, --plan-only, --full, --variant, --allow-fetch, --ask-format, --list-refs). Handle lock … as … as its own command.
  2. Classify remaining words that exactly match a known form, domain (incl. 3d / webgl), or modifier.
  3. Join every leftover word, in order, as the brief. Empty brief is allowed. Multi-word briefs are the point.
  4. A known modifier still loads its pack (neon, handdrawn, …). Leftover words around it stay in the brief.

The brief is a lens, not new content and not a closed theme list. Apply it to ground, motif, pattern, type, and motion. Source facts stay source facts. Log it on the report Brief: line (or none).

0. Context sniff (do not skip)

In parallel, cheaply:

  • Workspace root, README, GOAL.md if present (read; do not overwrite)
  • git status -sb if a repo
  • Open / recently viewed files if known
  • User text after the slash
  • Domain files the repo already points at (CONTRIBUTING, protocol docs, failing tests)
  • The target file if the user pointed at one (prefer .html; if they pointed at PDF/PPTX/DOCX, wrap or extract into HTML before calling the engine)

One-sentence lock: what this is, what happens to them.

If context is empty (blank chat, empty folder) and no interview: still ship. Invent from the folder name and any README. Do not stall for a brief.

0.5 Interview category (skip unless chosen)

Only when they passed interview. You choose the category; the agent decides the questions.

  • After the sniff, ask only what context cannot answer.
  • One question at a time. Wait. Multiple questions at once is bewildering.
  • Each question includes a recommended answer. They can take it, replace it, or say just go to skip the rest and build.
  • Questions come from the note bank (device / leap / craft / effect) — not "what style" or "any preferences?"
  • If the codebase can answer, Read/Grep instead of asking.
  • Cap 4 questions, then build. Do not leave the session in interview forever.
0.75 Adjacent possible (private)

From the parts already in this repo or thread, name one unused combination. That is the leap candidate. Do not dump a mood board.

Pick one SCAMPER letter as the mutation (Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse). Do not run all seven in chat.

0.85 Anchor list (private, but every plate must map to one)

From the source, extract 3–5 concrete anchors: proper nouns (people, places, brands), dates, numbers, verbs of action, physical objects, quoted phrases. These are the units of content the design must serve.

Every plate, tile, section, chart, or hero unit in the output must map back to at least one anchor. If it does not, delete it — do not paint a placeholder tile with the word blank, sample, or an alt-text stand-in (see must-not).

Record the anchor → plate mapping privately (or in the local ledger). Cite it on the report Anchors: line so the user sees the design isn't drifting from the source.

Example — Texas notebook source: anchors = {Jordan Rivers, Big Bend, 1836, Post Office, west Texas sunset}. Draw C's shader-hero → sunset; letterpress-card #01 → 1836; map plate → Big Bend; caption → Rivers; kicker → Post Office. Every unit serves an anchor.

1. Four notes (private)

From references/notes.md: one Device, one Leap (vastness + accommodation, moral beauty, big ideas, small self), one Craft, one Effect. Mutations name a cut, plate, magnet, withheld title, first-run beat, API shape, or demo — not "cinematic."

Do not dump the bank in chat.

2. Form router

Follow references/forms.md unless a category forced the family.

Webpage / HTML / infographic → load references/webpage-craft.md before writing the file. If the user added a domain (second word), also load the matching pack in references/domains/. Token infographic — as a visual form or as a domain — always loads references/domains/infographic.md and references/research/infographic-craft.md. Lock structure (list / sequence / hierarchy / compare / values / relation) from this source before drawing marks. If they added a modifier (third word or --style), also load the matching pack in references/modifiers/. Leftover words are the brief (step 0) — follow them; do not require a pack file. If they passed --ref <name>, load references/locks/<name>.md instead of (or in addition to) a domain — treat locks as domain packs.

SVG → load references/forms/svg.md. Marks in the field, type in the gutter. No label on a path. Alive-micro default: 2–4 loops on marks (weenie-breathe + one fact-loop + hover pairing). Brief still / no-motion / print freezes loops. Leftover words are a lens. Do not clone the Texas gold weenie onto a different source.

3js → load references/forms/3js.md. Reserved HUD strip. Silhouettes from this source. Not a primitive demo. Alive-micro default: idle life on 2–3 meshes (weenie-turn, flow along a source curve, sun-breath) plus camera ease / wide-drift. Brief still pins the camera and freezes meshes. Leftover words are a lens. Do not clone the Texas gold room onto a different source.

simulation → load references/forms/simulation.md. Playable model of this source's sequence (years, days, handshake steps — whatever it times with). Type in the gutter. Marks on the field. Nested short spans are inspectable. Default paused on the first fact. Leftover words are a lens. Do not clone the Texas 1836–1995 clock onto a different source.

PDF / document / slides → host conversion only. Prefer: extract or wrap as HTML, run npx reimagine-it, ship HTML. Load references/forms/ only if the user explicitly asked for that format and Weasyprint / python-docx / python-pptx (etc.) is on the machine. Do not present those packs as CLI tokens.

Any webpage output — with or without a domain / modifier / lock — must land hero-scale inline SVG doing real work, three moving elements at any moment, and 3D that reads in a still (rotation ≥ 12° + shadow blur ≥ 24px, or translateZ ≥ 30px + real shadow, or inline WebGL2). If a screenshot cannot prove all three, the redesign did not earn the form. Exception — infographic: the poster stays orthographic (paper drop-shadow only; no rotateX on the board).

Non-HTML export — optional, never the default claim. If the user asked for PDF/PPTX/DOCX and the toolchain exists, follow the matching form pack. Otherwise ship HTML and name the missing tool.

Form follows the leap:

  • Capability in this codebase → implement here (plus a tiny proof: test, CLI, fail-demo).
  • Seeing-tool (infographic, weenie, canvas, explainer HTML, one-shot deck/PDF) → <workspace>/reimagined/<yyyy-mm-dd>-<slug>/.
  • Never hide a product feature inside reimagined/ as a souvenir.
2.4 Creative variation (never repeat)

/reimagine-it must not return the same output twice for the same source unless the caller pinned a seed with --seed <n> (or picked a named --variant). Every fresh run samples a new combination along the axes below so the same brief produces a real range of draws instead of one canned "official" answer:

AxisOptions (agent picks; content narrows the set)
Reader registerThe lens the whole page speaks through — governs pacing, voice, and how the other axes combine. dashboard-live (Bloomberg / product-page cadence), editorial-drift (magazine feature, deliberate pauses), field-guide-quiet (Colossal / Feixen still confidence), cinematic-shader (Cartier / Lando Norris — full-bleed shader hero, sound-optional), neubrutalist-blunt (thick borders, hard shadow, saturated palette, no gradients), poster-jazz-improv (Troxler — single type size, non-l-to-r reading, high asymmetry), data-humanist (Lupi / Fragapane — the identity IS the data, custom visual alphabet from the source).
Ground / palette weightingThe content-derived palette usually contains ~4 hues (e.g. Texas notebook → navy · cream · red · gold). Each draw picks a different anchor hue for the ground: deep-night (navy ground, warm accent), parchment (cream ground, red accent), void (near-black ground, single-hue accent), raw-paper (off-white ground, ink accent), field-blue (mid-blue ground, cream accent), shader-glow (fullbleed animated gradient/shader as ground, ink on top).
Hero movekpi-skyline, illustrated-map, kinetic-type-headline, inline-shader-hero (WebGL2 full-bleed with scroll-driven uniforms), oversized-numeral, letterpress-plate, photograph-strip, weenie-object, variable-font-morph-hero (scroll-driven wght/wdth axis pulse; letter-spacing buffer reserved), sticky-evidence-pin (BBC-Lost-Tablet pattern — artifact stays fixed while narrative scrolls), oversized-quote-with-drop-cap, priestley-timeline (dates on a common year scale), isotype-unit-count (N copies of a same-size pictogram), portrait-grid-poster (hero encoding + supporting grid).
Infographic structure (when form or domain is infographic)Sniff this source first (AntV-style router, not their templates): list (unordered peers), sequence (steps / dated beats), hierarchy (tree), compare (two+ groups), values (magnitudes), relation (nodes+edges the source actually names). Then pick InfoAlign scenery: grid, star, portrait, landscape, portrait-grid, spiral (spiral only if the metaphor is a spiral). Gold Texas is sequence+compare on portrait-grid; gold Jules is sequence+star around the cone. Do not default to either.
Plate styleFor a 3-item section: dashboard-tile, editorial-dropcap, letterpress-card (numbered, with stamp), line-art-token, photograph-plate, bento-cell, index-card-stack, custom-data-glyph (Fragapane-style organic mark derived from a datapoint — leaves, braids, snakes, whatever the content demands), poster-tile-one-size (Troxler — every info at one type size, composition alone carries hierarchy).
Motion budgetdashboard-live (counters + pulses), editorial-drift (petals / dust / paper), kinetic-type-sway (headline sways), shader-loop (fullbleed shader), still-with-one-loop (one animated element), no-motion, scroll-driven-axis-morph (variable font wght/wdth bound to animation-timeline: scroll(root)), sticky-highlight-reveal (BBC / Pudding — pinned evidence with contextual line lighting), view-transition-morph (@view-transition { navigation: auto; } for multi-page packs — zero-JS native crossfade). When the hero form is svg or 3js: force alive-micro (2–4 fact-tied loops + hover pairing) unless the brief says still / no-motion / print. When the hero form is simulation: the motion is the clock (day-scale then years); default paused on the first year; do not sample webpage motion onto it. Do not sample a webpage-only motion (shader-loop, view-transition-morph) onto a weenie, a Three.js field, or a clock.
Type accentsans+mono, serif+italic, small-caps, mixed-italic, display-cut (oversized cuts), blackletter+grotesk, custom-display+workhorse-sans (M/M Paris pattern — one bespoke display voice + one neutral sans body — Editorial New + Neue Montreal, or Söhne + Signifier, etc.), variable-single-family (one variable font family, three axes doing the hierarchy work — GT Standard / Recursive).
3D signaturecard-fan (rotateY ±14°), letterpress-deboss (inset shadow), floating-hero (translateZ 40px), depth-strata (three z-layers), parallax-scroll, no-3D-just-shadow, matcap-hero (Bruno Simon technique — matcap texture on a real Three.js mesh for fake-but-convincing shading with zero light math), alcove-scenes (Cartier — a handful of self-contained 3D scenes each keyed to a section, hidden gestures reward exploration).

Rules:

  1. Never repeat the previous draw's exact combination. Track it in memory or the local ledger for the session.
  2. Content narrows the set — do not pick card-fan for a printed field guide, do not pick letterpress-deboss for a WebGL cinematic. The content decides which sub-space is coherent; variation happens inside it. The leftover brief may reweight that sub-space (ground, pattern, type, motion) without adding facts.
  3. The reader register governs coherence. Once you pick a register, the other axes must fit its grammar. A cinematic-shader register needs inline-shader-hero + shader-loop motion + matcap-hero or alcove-scenes 3D; a poster-jazz-improv register needs poster-tile-one-size plates + no-3D-just-shadow + high asymmetry. Do not glue incompatible axes. svg / 3js forms ignore webpage motion samples — they use alive-micro from the form pack. simulation ignores them too — it uses the clock in forms/simulation.md.
  4. --seed <n> pins the sample deterministically. Given the same source + same seed, the output must be byte-equivalent so users can reproduce a specific draw for locks or PRs.
  5. --variant a / --variant b / … are named seeds. a is the first canonical draw, b the second, etc. Ship them under after.html, after-2.html, after-3.html when a gold pack demonstrates the variance.
  6. Show the sample in the report Draw: line so the user knows which combination was picked (e.g. Draw: cinematic-shader · shader-glow · inline-shader-hero · custom-data-glyph · scroll-driven-axis-morph · variable-single-family · matcap-hero).

Gold demonstration (gold/webpage/): Draw A after.html (dashboard-live · deep-night), Draw B after-2.html (editorial-drift · parchment), Draw C after-3.html (cinematic-shader · shader-glow — the raised bar from v2.2 research). twins.png proves the range at a glance.

Gold infographic (gold/domains/infographic/after.html, v2.3): one Texas-notebook draw. Live runs derive palette, pattern, glyphs, and composition from this source (plus leftover brief). Cloning that gold’s scenery or its poster chrome onto a different source is a fail.

Gold Jules (gold/jules/): one parlor draw. Default webpage is the parlor you walk into (full-bleed counter + cone), not a stationery slip. Artistic is a painting of this shop's materials (not gold's cream/coral sine waves). Photography is still-lifes of marble / freezer / walk-up (not a VOL/ISSUE folio clone). Infographic is a star around the cone. SVG is stacked scoops. 3js camera sits on the counter. Simulation is the flavor board. Proof that a second source must not look like the first.

Gold forms (gold/forms/): one Texas-notebook draw each. Live svg / 3js / simulation runs derive weenie, meshes, clock unit, geography, palette, and composition from this source. Do not reuse weenie-left / schematic-center / gutter-right / timeline-bottom, The years run, or an All three place HUD.

Show full SKILL.md (1,831 more words)Show less
2.5 Modifiers · font · lock (extend the pack)
  • Modifiers are additive and composable. glassmorphism waives the spine's "blur / glassmorphism as the design" cut-list entry and adds its own non-negotiables (real depth behind the glass, layered panels, reduced motion budget). bento restructures the section grid into named tiles. Modifiers stack (artistic glassmorphism is a real combination); packs must not fight the spine on grid / palette cap / one motif.
  • Font override (--font "...") replaces the display or body family. Build a full CSS stack with sensible fallbacks (a serif family gets serif at the end, a mono family gets monospace). Never fetch a webfont; if the user wants one, they must pass --allow-fetch and understand it breaks the offline promise.
  • Lock: on /reimagine-it lock <path> [as <name>], read <path> (HTML/CSS/JSON/PDF metadata/etc.) and extract palette + type stack + motifs + motion + 3D signatures + section structure. Write references/locks/<name>.md as a full pack. Later /reimagine-it <target> --ref <name> loads that pack as if it were a domain, so the same design DNA can be applied to a different target (or a different medium — a webpage lock can inform a slides pack).
2.6 Output format(s) — HTML is the engine product

The CLI and playground always ship standalone HTML. That is the valid product claim.

  1. Default: run npx reimagine-it (or --auto) on HTML and write .html. Do not invent a PDF/PPTX/DOCX twin.
  2. If the source was not HTML: convert or wrap it as HTML first (extract text into a simple page). Then run the engine. Say so in the report Formats: line.
  3. If the user explicitly asked for PDF / slides / docx and the local toolchain exists, follow the matching pack in references/forms/ after the HTML artifact exists. If the toolchain is missing, ship HTML and name the exact next command — do not claim the engine wrote that format.
  4. --ask-format is only for that optional export step, not for choosing whether the engine works. Default remains HTML.

Log the decision on the report Formats: line (usually html).

3. Build

Include a 5-line README.md next to one-shot folders: what it is, how to run/open, which note drove it. In-place code changes: the proof (test or demo command) is the README.

Hero craft (any form):

  • One magnet in the first encounter (weenie, first-run command, first sentence, first failing-then-green demo, first spread, first slide)
  • Real content from this context — no lorem, no fake stats, no invented APIs
  • Effect before method: they should be able to say what happened, not only how you did it
  • Withhold the label until the artifact has done work

Stretch (required in the report): one thing they did not know was in bounds. Build when it is one extra file or a small sibling; otherwise give the exact next slash (/reimagine-it auto, /reimagine-it webpage cinematic, /reimagine-it lock <path>, …).

4. --full plus-pass

After the hero exists: re-read it. Apply one plus (criticism that contains a new move). Do not restart. Do not add a second product.

4.5 --variants N (optional)

If the user asked for N variants, produce N distinct outputs from the same brief. Each variant must land a different make-strange move (not the same page with a new color). Write to <hero>/var-1/, <hero>/var-2/, … and ship a strip.png composite so the user can pick.

5. Verify (functional + visual)

Two passes. Skipping either is a partial at best, not a shipped.

5.a Functional pass — the artifact does the thing:

  • Path exists. Report it as a markdown link.
  • Code/CLI: the command or test actually ran; paste the exit or the proving line.
  • Visual file well-formedness: viewBox on SVG; HTML opens; no required CDN if offline was implied; PDF opens in a viewer; docx/pptx opens in the target app; ebook (mobi/epub) opens in Kindle Previewer / Calibre.
  • Motion / 3D: two stills 500 ms apart show visible pixel change; at least one element has ≥ 12° rotation + ≥ 24 px shadow blur, or inline WebGL2.
  • Prose: the piece is a file they can keep, not only chat.
  • Protocol: a spec they can implement or a spike that runs — not only a metaphor.

5.b Visual verification pass — actually look at the render.

Render the hero into an image (headless Chrome for HTML → PNG at ≥ 1400 px wide; PDF → first-page PNG; docx → export to PDF then PNG; pptx → first-slide PNG; mobi/epub → open in Previewer and screenshot the first two pages). Read the image tool result (or open it in the IDE viewer) and manually scan for every one of these failure modes before reporting shipped:

  • Blank plates / placeholder labels. No visible element may literally read blank, placeholder, TBD, TODO, lorem, …, […], xxx, sample text, Title goes here, caption, or an alt-text stand-in. If a slot has no real content from the source, delete the slot — do not paint a card with the word "blank" on it.
  • Every plate maps to an anchor. Cross-check every visible unit against the anchor list from step 0.85. Unmapped tiles are drift — delete them.
  • Clipped / overlapping text. No label is cut off by another element (e.g. POST OFFICE rendered as POST O CE because a foreground shape overlaps the text). Fix z-index / padding / overflow or move the overlapping element. SVG / 3js: no label sits on a mark, mesh, or path; 3js HUD is a reserved strip, never a stack on the canvas.
  • Broken image / broken svg. No alt text is showing where a picture should be. No <svg> renders empty.
  • Runaway columns / squashed hero. Nothing extends past the viewport. Hero is not vertically flattened.
  • Off-palette accent. Every colored element is on the content-derived palette. No stray CSS-default blue link, no browser-default <button> chrome.
  • Wrong content. All copy on the render actually appears in the source (or is a caption/index the skill added). No fabricated place names, dates, statistics, or people.
  • Motion proof. If the pack claims motion (svg/3js default is alive-micro), capture two frames (~600 ms apart) and compare hashes; identical hashes = motion did not run. Brief still is the exception.

5.c Craft-floor pass (webpage only) — grep the produced HTML/CSS and confirm the craft-floor contract holds: Run the deterministic audit: npx reimagine-it audit <output-file.html> (or /reimagine-it audit). This runs 19 checks across typography, palette, motion, content, structure, and performance with no LLM and no API key. Python scripts/audit.py is the GitHub Action mirror and must agree. Zero failures must pass; warnings are advisory. Supports --verbose and --json for CI.

  • Focus & selection: at least one ::selection rule; at least one :focus-visible rule; no outline: 0 / outline: none without a real replacement.
  • Reduced motion: a @media (prefers-reduced-motion: reduce) block is present and decomposes correctly (transitions off, focus indicators still visible, essential state feedback preserved).
  • Compositor-only motion: no @keyframes or transitions animate top, left, right, bottom, width, height, margin, padding, font-size, letter-spacing, line-height, word-spacing, color, or background-color (last two are OK for :hover on a static element only). Everything animated during scroll/interaction uses only transform, opacity, or filter.
  • No transition: all. Explicit properties, always.
  • Kinetic type reserves space: any variable-font axis morph has a letter-spacing buffer or min-width reservation on the animated element.
  • Interactive elements have hover + focus states (not just :hover).
  • Sound off by default: if --sound was not passed, there is no <audio autoplay>, no Howler autoplay: true, and no unmuted <video> outside an explicit user gesture.
  • Auto source safety: auto may read and write only the generated artifact/report; it must not mutate the source unless the user explicitly asks for an edit.

If any of these fail, patch the CSS in one pass; if the fix crosses into "add a whole subsystem," downgrade to partial and name the specific miss.

5.d Named-object + clone pass — load references/review.md. If the source names a public object (the Lone Star flag, a cone, a press), the weenie is that object — not a gold-star / cream-stripe remix. A second source must not wear Texas gold DNA or layout chrome. Gold fixtures: python scripts/review_gold.py.

Log the verify passes on the report:

  • Visual verify: with what you scanned for and what you fixed
  • Craft floor: with the results of the grep (e.g. ::selection ✓, :focus-visible ✓, prefers-reduced-motion ✓, compositor-only ✓, no transition:all ✓, sound off ✓).

If any failure mode is present and cannot be fixed in one pass, ship partial and name the specific bug — never dress a placeholder up as done.

6. Report
REIMAGINED: shipped | partial | blocked
Mode: reimagine-it
About: <one sentence>
Hero: <path + how to run/open>
Domain / modifier / --ref: <if any>
Brief: <leftover phrase, or none>
Font stack: <if --font was passed>
Anchors: <3–5 nouns/proper nouns/dates from the source that every plate mapped back to>
Draw: <reader-register> · <ground> · <hero-move> · <plate-style> · <motion> · <type-accent> · <3D-signature>
      (svg / 3js: motion is `alive-micro` unless the brief was `still`)
Seed: <n if pinned, else "fresh">
Formats: <usually "html"; name any optional host export>
Sound: <off | ambient | feedback | full — with tier + mute-control + no-autoplay confirmed>
Stretch: <what they didn't know was possible>
Notes: <only if --notes>
Functional verify: <what you actually ran, opened, or checked>
Visual verify: <scan result: no blank plates? no clipped text? every plate maps to an anchor? palette on-source? named weenie is the real object (flag cloth is white star / white over red, not gold-on-parchment)? motion advanced? composition is this source's object, not gold's layout chrome?>
Craft floor: <::selection ✓, :focus-visible ✓, prefers-reduced-motion ✓, compositor-only motion ✓, no transition:all ✓, sound off unless --sound ✓>

Lead the user-facing reply with the artifact and the stretch, not the protocol.

Must not

  • Ship a bullet list instead of an artifact
  • Return the same draw twice for the same source without an explicit --seed/--variant pin
  • Drop leftover user words because they are not a named form, domain, or modifier
  • Clone gold DNA or gold composition onto a source that is not that notebook — not the palette (Texas parchment/navy/star-red), not the scenery (Lone Star, Alamo chapel, 1836–1995 map-clock), and not the layout chrome (00 · MASTHEAD rail, /reimagine-it infographic kicker, weenie-left / map-center / gutter-right / timeline-bottom SVG, The years run + All three HUD). Method travels. Composition does not.
  • Report shipped without the visual verification pass (5.b) — no exceptions
  • Paint a plate that literally reads blank, placeholder, TBD, TODO, lorem, sample, caption, …, […], Title goes here, or any alt-text stand-in. Empty slot → delete the slot. Real content only.
  • Ship a render with clipped or overlapped text (e.g. a foreground shape covering half a label). Fix z-index / padding / overflow before reporting shipped.
  • Hand-author a redesigned page when npx reimagine-it can run. The CLI is the engine; the skill is the wrapper.
  • Claim the CLI redesigns PDF, PPTX, DOCX, MOBI, or “any file.” HTML in, HTML out. Host conversion is extra and optional.
  • Silently pretend a missing export toolchain ran. If the user asked for PDF/PPTX and the tool is missing, ship HTML and name the next command.
  • Ship a webpage output that fails the craft floor (§5.c) on any of: ::selection styled, :focus-visible styled, prefers-reduced-motion block present and decomposed correctly, compositor-only motion, no transition: all, no outline: 0 without a real replacement, no autoplay sound unless --sound was passed. Patch in one pass or downgrade to partial.
  • Animate top / left / right / bottom / width / height / margin / padding / font-size / letter-spacing / line-height / word-spacing / color / background-color for anything that runs during scroll or interaction. Use transform, opacity, or filter only — anything else forces layout recalc every frame and fails visual verify.
  • Use transition: all — explicit properties always. all thrashes on any parent change.
  • Ship a page with outline: 0 / outline: none and no visible focus replacement — WCAG 2.4.7 non-negotiable.
  • Paint a plate that does not map back to an anchor from step 0.85 (i.e. drift into invented content). Every visible unit serves a source anchor or it is deleted.
  • Treat /reimagine-it as graphics-only
  • Let auto overwrite the source or hide which direction was selected
  • Interview without the interview category
  • Ask "what style do you want?"
  • Use wow, "make it pop", or shock-as-strategy
  • Call paid image/video APIs without asking
  • Clone Dribbble / Collect UI as the idea
  • Fake calligraphy or fake stats
  • Fabricate content the source does not contain (invented place names, made-up statistics, phantom people)
  • Scaffold a greenfield app into an unrelated repo without being asked
  • Fetch a webfont without --allow-fetch
  • Save a lock outside references/locks/ (or the host's configured locks path)

© Kayforkind, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 37 other files (references) in skills/reimagine-it of Kayforkind/reimagine-it.

  • SKILL.md
  • examples.md
  • references/craft-floor.md
  • references/domains/README.md
  • references/domains/artistic.md
  • references/domains/cinematic.md
  • references/domains/dashboard.md
  • references/domains/ecommerce.md
  • references/domains/infographic.md
  • references/domains/landing.md
  • references/domains/photography.md
  • references/domains/portfolio.md
  • references/forms.md
  • references/forms/3js.md
  • references/forms/README.md
  • references/forms/document.md
  • references/forms/pdf.md
  • references/forms/simulation.md
  • … and 20 more

Open the folder on GitHubat commit e8c9378

Compare with similar skills

Reimagine It next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Reimagine It compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Reimagine It this skillKayforkind/reimagine-it205—~11kAutomated safety check: PassMIT
MarkitdownImCa0/just-laws78114 repos~3.2kAutomated safety check: NotesMIT
GenOffice Document CLIgenspark-ai/genoffice9.2k—~19kAutomated safety check: PassApache-2.0
Docx4jplutext/docx4j2.4k—~2.5kAutomated safety check: PassNone
Cyber Pptcrazyykhllc-bit/CyberPPT1.8k—~10kAutomated safety check: PassMIT
Docsagentdocsagent/docsagent625—~834Automated safety check: PassNone

Similar skills

  • Markitdown

    ImCa0/just-laws

    Convert files and office documents to Markdown. An agent skill from ImCa0/just-laws.

    781 GitHub starsUsed in 14 repos~3.2k tokens
    Documents & OfficeAuto-check: notes
  • GenOffice Document CLI

    genspark-ai/genoffice

    Creates, converts, reads and edits real pptx, xlsx, docx and PDF files locally through the genoffice command line.

    9.2k GitHub stars~19k tokensUpdated yesterday
    Documents & OfficeAuto-check passed
  • Docx4j

    plutext/docx4j

    A skill your agent uses when writing Java code that creates, reads or edits Word (.docx), PowerPoint (.pptx) or Excel (.xlsx) files with docx4j — including generating documents, editing existing…

    2.4k GitHub stars~2.5k tokensUpdated 2 days ago
    Documents & OfficeAuto-check passed
  • Cyber Ppt

    crazyykhllc-bit/CyberPPT

    当用户需要把 DOCX、PDF、TXT、XLSX、研究报告、业务材料或原始数据转成高密度、可编辑、咨询风格 PPTX 时使用;也适用于需要 SCR 论证、视觉风格探索、详细图表和渲染质检的 PPT。

    1.8k GitHub stars~10k tokensUpdated 2 mo ago
    Documents & OfficeAuto-check passed
  • Docsagent

    docsagent/docsagent

    Search and manage private, local document collections (PDF, PPTX, DOCX) offline.

    625 GitHub stars~834 tokensUpdated 12 days ago
    Documents & OfficeAuto-check passed
  • Markit

    shift-labs-ai/markit

    Convert files and URLs to Markdown. An agent skill from shift-labs-ai/markit.

    1.3k GitHub stars~299 tokensUpdated 1 mo ago
    Documents & OfficeAuto-check passed

More from Kayforkind/reimagine-it

  • Reimagine It Audit

    Kayforkind/reimagine-it

    Design Health — runs 19 deterministic quality checks on HTML output.

    205 GitHub stars~756 tokensUpdated 4 days ago
    Auto-check passed
  • Reimagine It Extract

    Kayforkind/reimagine-it

    Emit the content signals reimagine-it reads from an HTML file — title, anchors, proper nouns, dates, numbers, emails, links, source hex colors, and the derived palette — as JSON without generating a…

    205 GitHub stars~819 tokensUpdated 4 days ago
    Auto-check passed
  • Reimagine It Generate

    Kayforkind/reimagine-it

    Generate one Content-Derived Design direction from an HTML source, or let Auto choose the strongest direction.

    205 GitHub stars~876 tokensUpdated 4 days ago
    Auto-check passed
  • Reimagine It Lock

    Kayforkind/reimagine-it

    Design DNA capture — extracts palette, type stack, motifs, motion, and 3D signatures from a shipped HTML file and saves them as a reusable reference pack.

    205 GitHub stars~959 tokensUpdated 4 days ago
    Auto-check passed
  • Reimagine It Variations

    Kayforkind/reimagine-it

    Generate several distinct design directions from one HTML source plus a contrast sheet that shows every direction side by side.

    205 GitHub stars~680 tokensUpdated 4 days ago
    Auto-check passed
  • Reimagine It Infographic

    Kayforkind/reimagine-it

    Content-derived infographic poster — reads the source file and builds a statistical poster from the facts already in it.

    205 GitHub stars~1.1k tokensUpdated 4 days ago
    Auto-check passed

Questions about Reimagine It

What does Reimagine It do?

Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file. Reimagine It is an agent skill from Kayforkind/reimagine-it. Content-Derived Design — reads existing HTML and writes a stronger standalone page from headings, facts, names, dates, numbers, links, emails, and colors already in that file.

When should I use Reimagine It?

Reimagine It fits situations like: the user says /reimagine-it; redesign this page; make an infographic; wants a content-aware redesign.

How do I install Reimagine It in Claude Code?

Run `npx skills add Kayforkind/reimagine-it --skill reimagine-it -a claude-code`. Or copy the skill folder (skills/reimagine-it in Kayforkind/reimagine-it) into .claude/skills/reimagine-it in your project. Claude Code loads it when a task matches its description.

How do I install Reimagine It in Codex?

Run `npx skills add Kayforkind/reimagine-it --skill reimagine-it -a codex`. Or copy the skill folder (skills/reimagine-it in Kayforkind/reimagine-it) into .agents/skills/reimagine-it in your project. Codex loads it when a task matches its description.

Can I use Reimagine It in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Kayforkind/reimagine-it --skill reimagine-it -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/reimagine-it, .gemini/skills/reimagine-it, .github/skills/reimagine-it and .opencode/skills/reimagine-it in your project.

What does Reimagine It need to run?

Going by SKILL.md and its folder, Reimagine It needs the command-line tools its instructions call (npx, git, just and python). Our summary lists: Python 3; Node.js.

Does Reimagine It access the network?

SKILL.md contains no URLs. Its commands use npx and git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Reimagine It safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Reimagine It use?

Reimagine It is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Reimagine It use?

About 11k tokens (SKILL.md is roughly 44k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 46k tokens, read only when the agent opens those files.

What are the alternatives to Reimagine It?

Skills that share tags, products or a category with Reimagine It: Markitdown (ImCa0/just-laws, 781 stars), GenOffice Document CLI (genspark-ai/genoffice, 9.2k stars), Docx4j (plutext/docx4j, 2.4k stars) and Cyber Ppt (crazyykhllc-bit/CyberPPT, 1.8k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Reimagine It?

Kayforkind (a GitHub user) maintains it in Kayforkind/reimagine-it, which has 205 GitHub stars. The repository holds 7 skills in this directory. The repository was last updated on October 6, 2026.

Source: Kayforkind/reimagine-it on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.