SureForge
SkillAn instruction-only skill that makes agents research before they ask, plan before they build, and verify before they call anything done.
- workflow
- verification
- quality
- skill
An agent skill and zero-dependency npm CLI that reviews your app's UI for the tell-tale signs of AI-generated design — 31 detector rules distilled from a study the README counts at 5,008,684 app reviews
unslopovich finds the UI patterns that make an app look AI-generated. The README's study line: 5,008,684 app reviews read, 1 purple gradient found, 31 detector rules, one command — and research/README.md publishes the study behind it, with the method and the code. Run npx unslopovich and the installer finds the coding agents on your machine and installs for each one: Claude Code gets ~/.claude/skills/unslopovich (/unslopovich review my app), Codex, Cursor and OpenCode get ~/.agents/skills/unslopovich, Gemini CLI and GitHub Copilot get per-project skill folders (.gemini/skills, .github/skills); each with the invocation shown in the README table. Scope it with --agent codex,cursor (any of claude, codex, cursor, opencode, gemini, copilot), commit a shared copy with --project, point other SKILL.md-reading agents at it with --dir, or install as a Claude Code / Codex plugin via the marketplace commands. The skill is SKILL.md, a reference/ folder it reads on demand, and the detector in scripts/. Needs Node 18 or later.
Use before shipping anything an agent built: a pre-release pass over the UI it generated, a sweep of a vibecoded internal tool before real users see it, or a design-review gate in CI via the --project committable folder. The README's own boundary applies: it reads code and (when your agent can open screens) looks at screenshots — it can't see your app run, can't profile lag or battery, won't fix your bugs, and taste is still yours.
One command installs for every detected agent (needs Node 18+): npx unslopovich Per-agent install targets: Claude Code ~/.claude/skills/unslopovich, Codex and Cursor and OpenCode ~/.agents/skills/unslopovich, Gemini CLI .gemini/skills/unslopovich (per project), GitHub Copilot .github/skills/unslopovich (per project). Then type /unslopovich review my app (Claude Code, Cursor, OpenCode, Gemini, GitHub Copilot), $unslopovich review my app (Codex), or just ask in your own words. Selected agents only: npx unslopovich --agent codex,cursor. Team-shared copy: npx unslopovich --project and commit the folder. As a Claude Code plugin: /plugin marketplace add erkamyaman/unslopovich then /plugin install unslopovich@unslopovich.
The headline numbers are the project's self-reported study (5,008,684 reviews, 31 rules) — the study's method and code are published in research/README.md, but they are the author's own counts, not an independent audit. The before/after screenshot in the README is labeled a dramatization. Rule provenance is reviews and platform guidelines, so taste disagreements are expected — the README's what-it-won't-do section is candid: it can't see your app run (the detector reads code; the skill looks at screens only when your agent can open them and says so when it couldn't), it can't profile lag or battery (reports say Not profiled), and it won't fix your bugs. MIT. 233 stars as of writing (2026-11-19).
Repository
erkamyaman/unslopovichCharted
An instruction-only skill that makes agents research before they ask, plan before they build, and verify before they call anything done.
Guizang's agent skill turns your repo's recent updates into a finished promo video — real components, code-rendered animation, an original code-written soundtrack — no video editor involved.
An agent skill that geolocates a photo with nothing readable in it — OSM geometry, elevation skylines and satellite tiles turn “where was this taken?” into a measured answer with an error radius.