unslop
Skill4 starsChanged 5 months agoMITView on GitHub
Remove the "vibecoded" / AI-slop look from an arbitrary frontend codebase at the ROOT, autonomously, and prove the result clears a crafted floor. Full-repo traversal (any framework) → root-cause attribution (the one file that declares each default: font, icons, color/radius/shadow tokens, gradient/glass, motion, copy voice) → fix once at the root, never at forty call sites → the substance tells no detector covers (legal routes, loading/empty/error states, placeholder and fake content, real product evidence) → render at 1280+390 → deterministic gate (scripts/unslop_gate.py, exit 0 = clear). Composes the impeccable detector (64 rules) and commands, Anthropic frontend-design for direction, broomva-design as the house example; reimplements none of them. USE WHEN: a site or app "looks vibecoded", "looks AI-generated", "looks like every other AI site", "feels generic/templated/slop", "make it feel premium / crafted / intentional", "de-slop / unslop this", "does this pass the vibecoded test", or before shipping any LLM-built UI. NOT FOR: designing a new surface from scratch (impeccable new-work / frontend-design), a11y/perf audits alone (impeccable audit), backend-only repos, or a one-line style tweak.
197 other projects use a version of this.
Adopted by
Most recently taken up by
- alastairlundy/skillstheir own version, first seen Nov 8, 2026
- sanjaibalajee/skillstheir own version, first seen Nov 8, 2026
- ku5ic/claude-kittheir own version, first seen Nov 8, 2026
- augchan42/inkstonetheir own version, first seen Nov 8, 2026
- caidanw/skillstheir own version, first seen Nov 8, 2026
- runkids/my-skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use principle-never-block-on-the-human, greploop-apps and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
jaydubya818Skillprinciple-never-block-on-the-human
Apply when safe reversible implementation details can proceed inside scope.
Also in 45 projects that use this
jaydubya818Skillgreploop-apps
Run greploop through @greptile-apps when the normal trigger rejects a large PR. Use only with the companion greploop skill.
Also in 5 projects that use this
ianclearySkillprinciple-explain-the-number
Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.
Also in 30 projects that use this
bnemaSkillsetup-pstack
Configure which Pi models pstack uses per role. Detects available provider/model specs and writes pstack's user configuration. Use for /setup-pstack, "configure pstack models", or changing pstack model choices.
Also in 26 projects that use this
ianclearySkillpoteto-help
Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.
Also in 15 projects that use this
- +7 more from jaydubya818
Other versions197 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 197
- tester-army/e2e★ 13ksame name
- heydamir/unslop★ 12ksame name
- jaredpalmer/kev★ 9.8ksame name
- backnotprop/pstack★ 3.8ksame name
- michaelshimeles/rstack★ 1.4ksame name
- michaelshimeles/skills★ 1.4ksame name
- VladSez/easy-invoice-pdf★ 1.1ksame name
- Silo-Server/silo-server★ 899same name
- moeru-ai/xsai★ 667same name
- nodetool-ai/nodetool★ 615same name
187 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository108 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/research/deep-dive-research-orchestrator/AGENTS.mdAGENTS.md
- skills/research/deep-dive-research-orchestrator/CLAUDE.mdCLAUDE.md
- skills/science/ocean-genomics/CLAUDE.mdCLAUDE.md
- skills/video/content-engine/CLAUDE.mdCLAUDE.md
- openrocket-simSkill
- sdr-satelliteSkill
- livecodingSkill
- omnivoiceSkill
- talkbackSkill
- drift-checkSkill
- morning-briefingSkill
- stakeholder-updateSkill
- weekly-reviewSkill
- d1-cliSkill
- procurerSkill
- swapitSkill
- agentic-vpsSkill
- colab-remoteSkill
- forkable-shellSkill
- remote-gpuSkill
- arcan-glassSkill
- brand-iconsSkill
- broomva-designSkill
- design-distillSkill
- design-engineeringSkill
- tektonSkill
- finance-substrateSkill
- haimaSkill
- investment-managementSkill
- wealth-managementSkill
- agentic-control-kernelSkill
- architecture-design-principlesSkill
- autoreviewSkill
- bstackSkill
- cross-reviewSkill
- dogfoodSkill
- harness-engineering-playbookSkill
- keelSkill
- legal-readinessSkill
- spec-contractSkill
- unhobbleSkill
- founder-mode-oncologySkill
- healthSkill
- bookkeepingSkill
- braindumpSkill
- ccrSkill
- colombia-conflictSkill
- comprehendSkill
- goodiesSkill
- kgSkill
- whatSkill
- claude-code-channelsSkill
- claude-remote-sessionsSkill
- bitnetSkill
- heretic-abliterationSkill
- tribe-v2-agent-alignmentSkill
- tribe-v2-bci-appliedSkill
- tribe-v2-neuroscienceSkill
- arcSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 193 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 61
Never hand control back mid-arc.
“Common” is measured: kept by at least 6 in 10 of the 193 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
unslop — remove the vibecoded look at the root, prove the crafted floor
Someone says "this looks vibecoded" or "make it feel premium and crafted with care". unslop runs
one arc over the whole codebase and ends with a PR whose gate is green or whose remaining reds are
named and handed to a human — never a cosmetic pass over the pages someone happened to open.
The one rule
Fix the class, not the example — and prove it on the rendered result.
A default (Inter, Lucide, purple-to-black,
rounded-2xl, glass cards, em dashes) is declared in one place and consumed in forty. Edit the one place. Then screenshot at 1280 and 390 and run the gate; reasoning is not validation. Twenty-six of the thirty tells are surface and a prompt removes them; the four that are absence of substance (no real demo, no loading states, no TOS, no privacy policy) are removed only by the real product — the gate keeps them red until a human makes them true.
Deterministic core (scripts/)
| Script | What it does | Exit |
scripts/unslop_survey.py <repo> [--detect] --json OUT --md OUT | Full-repo inventory: framework + routes; roots (each default → declaring file + blast radius + the single fix); copy tells; substance (legal, async-state coverage, placeholders, testimonials, pricing scaffold, product evidence, motion, design docs); impeccable findings if --detect | 0 / 2 usage |
scripts/unslop_gate.py <repo> [--manifest M | --detect] --evidence DIR [--waivers W] [--profile] [--strict] | The crafted floor as PASS/WARN/FAIL per check (direction.authored, detector.clean, fonts.deliberate, icons.single-system, tokens.*, copy.*, substance.*, motion.reduced-motion, evidence.render); waivers need a reason ≥20 chars | 0 clear · 1 FAIL · 2 usage |
Stdlib Python, no edits to the target, any framework (Next app/pages, SvelteKit, Astro, Nuxt, Remix,
Vite+React, static HTML). The detector is composed: it is found in the installed impeccable skill
or via npx impeccable detect; when absent the survey says unavailable and the gate WARNs — it never
pretends classes A–C were checked. Override with UNSLOP_DETECTOR="node /path/detect.mjs".
The arc (references/arc.md — read it before running)
0 Snapshot → 1 Survey → 2 Direction (decide once, DESIGN.md) → 3 Root plan
→ 4 Fix at root → 5 Structure + copy → 6 Class D (states · legal · evidence · truth)
→ 7 Render + inspect at 1280/390 → 8 Gate → 9 PR + reportNever hand control back mid-arc. Never ask "what style do you want?" — infer the direction from the product's own material and state it in DESIGN.md; if there is none, choose the quietest committed world and say so. Never fabricate substance (policy text, testimonials, demos) to turn a red green.
Composition map (bstack-native names; generic behavior in parentheses)
| Step | Composes |
| Survey / gate | scripts/unslop_survey.py, scripts/unslop_gate.py (this skill) |
| Mechanical tells (classes A–C) | impeccable detector detect --json <dir> (64 rules) + its per-edit hook |
| Direction | impeccable init / document / new-work; Anthropic frontend-design; broomva-design DESIGN.md as the house example of stated decisions |
| Root fixes | impeccable extract, typeset, colorize, layout, quieter; references/root-cause-playbook.md |
| Structure + copy | impeccable shape, distill, clarify, polish |
| Class D | impeccable harden, onboard + a human for what must be true |
| Render evidence | Playwright / agent-browser screenshots (bstack P11 Empirical) |
| Audit | impeccable audit (a11y / perf / responsive) — never re-implemented |
| Ship | bstack P4 Pipeline, P20 Cross-Review for design-system-touching diffs |
References
references/arc.md— the nine steps, with what each composes and the anti-rationalizations.references/crafted-floor.md— the operational definition of "crafted" (typography · spacing · color · motion · states · copy · a11y · perf · consistency · trust), every rule tagged [M] machine-checkable or [J] judgment, sourced (Kowalski, Freiberg, Linear, Comeau, Refactoring UI, Butterick, Rams, Apple, Material, GOV.UK, Vercel, Stripe, NN/g, Fogg, WCAG, web.dev), contradictions resolved explicitly.references/tells.md— the dated, regenerable list of current LLM defaults (the reel's 30 → class → machine check), Vercel's reject list, the no-ai-slop prose-pattern catalog (dated), and why a static ban-list decays.references/root-cause-playbook.md— per root kind: where the default lives, the one edit, the order.
Anti-rationalization
| Excuse | Reality |
| "The pages I opened look fine now." | Traversal is the point. Run the survey; roots[] and routes[] are the scope, not your tabs. |
| "I replaced Inter in every component." | You fixed the example. Re-survey: the old face's site count must reach zero by one edit at the root. |
| "The impeccable detector is clean, done." | Classes A–C only. The gate's other half is substance + render evidence; a clean detector with lorem and no privacy route is still slop. |
| "I'll generate a privacy policy / testimonials to pass the gate." | That turns an absence-of-substance tell into a fake-content tell. Leave it red, hand it to a human, name it in the PR. |
| "The tell list says no Lucide, so I'll switch icon sets." | The list detects, it doesn't ban. Keep it and state the decision in DESIGN.md; the gate reads it. |
| "It renders fine, I can tell from the JSX." | Screenshots at 1280 and 390 or it did not happen (P11). |
| "Should I ask which direction they want?" | No. Infer from the product's material, state it in DESIGN.md, proceed. A stated quiet choice beats an unstated loud default. |
Validation (skill self-test)
python3 -m pytest tests/green (survey + gate, both polarities, waivers, strict, evidence, CLI, and one regression per P20 finding — every case a false positive or fail-open a reviewer hit on a real repo).- Dogfood:
scripts/unslop_survey.py <a real repo> --detectproduces roots with aroot_file; the detector status isokwhen impeccable is installed andunavailable(never a silent pass) when not. unslop_gate.pyon the crafted fixture exits 0 with evidence; on the sloppy fixture exits 1 namingfonts.deliberate,substance.legal,substance.placeholders,motion.reduced-motionamong others.- skillify gate:
skillify_check.py skills/design/unslop --run-testsexits 0.
Provenance
BRO-2184. Origin: /checkit on an Instagram reel ("30 reasons your site looks vibecoded", 2026-08-15,
34k likes) → research/entities/pattern/vibecoded-tells.md (the 26-surface / 4-substance finding) →
this skill. Research synthesis: research/notes/2026-08-18-unslop-what-others-point-to-synthesis.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.

