anti-overengineering
Skill1 starsChanged 50 days agoApache-2.0View on GitHub
Enforces zero-dependency bias, native standard library utilization, anti-over-engineering, anti-YAGNI, and token-efficient lean architecture across all software projects and AI coding workflows. Trigger whenever proposing or installing third-party packages, designing architecture, creating helper utilities, refactoring, or generating code solutions. Mandates the 3-Gate Dependency Audit (Search before Import), Native Standard Library Replacement Matrix (Node/JS/TS, Python, Go, Web APIs), elimination of speculative enterprise abstractions (factories, excessive DTOs/interfaces for simple tasks), strict token conservation (diff-only output protocol), and proactive dependency de-bloating using the nobloat CLI engine.
- Installs packages
3 other projects use a version of this.
Adopted by
Most recently taken up by
- roronoazoroshao369/vibe-coding-ostheir own version, first seen Nov 15, 2026
- yueheng-rgb/codex-factorytheir own version, first seen Sep 15, 2026
- DGouron/review-flowtheir own version, first seen May 14, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- yueheng-rgb/codex-factory★ 100same name
- DGouron/review-flow★ 44same name
- roronoazoroshao369/vibe-coding-os★ 1same name
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 repository21 other files it gives its agents.
- min-tokSkill
- academic-scholarSkill
- agent-mcp-craftSkill
- api-security-reviewerSkill
- data-verifySkill
- deep-researcherSkill
- design-no-slopSkill
- enterprise-teamSkill
- git-craftSkill
- human-codingSkill
- laravel-architectSkill
- markdown-docSkill
- modern-stack-horizonSkill
- modern-web-guidanceSkill
- omni-pilotSkill
- perf-benchmarkSkill
- python-bug-fixerSkill
- react-ui-architectSkill
- reverse-engineer-anythingSkill
- typescript-nodeSkill
- tyw-auditSkill
The file
Files
Anti-Over-Engineering & Lean Software Craft (The Perfect Zero-Bloat Standard)
This skill forces the AI agent to operate with radical dependency restraint, architectural minimalism, and maximum token efficiency. It eliminates the reflexive AI habit of installing third-party packages for trivial tasks, building speculative enterprise abstractions, and exhausting context limits with verbose boilerplate dumps.
1. The Core Philosophy: Lean Engineering vs AI Bloat
AI models suffer from a statistical bias toward bloated open-source patterns—recommending outdated npm/pip packages from 2016 and wrapping 5-line functions inside multi-tier abstractions.
| Dimensi | AI Bloat Pattern (Over-Engineering) | Lean Engineering Standard |
| Ketergantungan Paket | Menginstal paket eksternal untuk fungsi 2 baris (misal: uuid, axios, rimraf, is-number). | Native Standard Library First: Menggunakan pustaka standar dan runtime modern. |
| Rantai Pasok (Supply Chain) | Mengabaikan beban transitive dependencies (1 paket menarik 40 sub-packages). | Transitive Blast-Radius Rule: Veto paket yang menarik $>3$ sub-dependencies untuk tugas sepele. |
| Arsitektur Solusi | Membuat 5 lapis abstraksi (Controller, Service, Repository, DTO, Factory) untuk CRUD sederhana. | KISS & Anti-YAGNI: Modularitas pragmatis, fungsi murni, dan struktur datar. |
| Konsumsi Token & Limit | Mencetak ulang seluruh file 500 baris hanya untuk mengubah 3 baris kode. | Diff-Only Output Protocol: Hanya keluarkan patch padat / targeted edit, hemat 80% kuota token. |
| Sikap Konsultasi | Mengiyakan tumpukan teknologi raksasa (Kubernetes, Kafka) untuk app kecil. | Tech-Stack Reality Check: Menantang over-engineering dan menawarkan solusi ramping. |
2. The 3-Gate Dependency Audit (Search & Verify Before Import)
Sebelum mengusulkan npm install, pip install, atau menambah dependensi baru, AI WAJIB mengevaluasi 3 Gerbang Verifikasi:
[Kebutuhan Fungsional]
│
▼
┌────────────────────────────────────────┐
│ Gerbang 1: The Native Capability Gate │
│ Bisa diselesaikan bawaan bahasa dalam │ ──(YA)──► GUNAKAN STANDARD LIBRARY / NATIVE
│ <= 25 baris kode? │ (DILARANG install paket baru)
└────────────────────────────────────────┘
│ (TIDAK)
▼
┌────────────────────────────────────────┐
│ Gerbang 2: The Existing Project Gate │
│ Apakah dependency sejenis SUDAH ADA di │ ──(YA)──► REUSE DEPENDENCY YANG SUDAH ADA
│ package.json / pyproject / go.mod? │ (DILARANG install pustaka tandingan)
└────────────────────────────────────────┘
│ (TIDAK)
▼
┌────────────────────────────────────────┐
│ Gerbang 3: The True Complexity Gate │
│ Apakah ini masalah tingkat lanjut? │ ──(TIDAK)─► TULIS LEAN HELPER FUNCTION MANDIRI
│ (Kriptografi berat, Parser AST, ML, DB)│
└────────────────────────────────────────┘
│ (YA)
▼
[IZINKAN DEPENDENCY EKSTERNAL DENGAN JUSTIFIKASI SINGKAT]Kriteria Larangan Keras (Strict Banned Packages):
- Helper Trivial:
is-odd,left-pad,is-number,lodash,underscore,ramdajika hanya butuh manipulasi array/objek standar. - Duplikasi HTTP Client:
axios,got,superagent,node-fetchjika runtime mendukungfetch(). - Pustaka UUID / Crypto Ringan:
uuid,nanoidjika platform menyediakancrypto.randomUUID(). - File System Utilities Jadul:
rimraf,mkdirp,ncpdi Node.js modern.
3. The 5 Pillars of Perfection
Pillar 1: Proactive De-Bloater (Auto-Purging Legacy Repos)
Bukan hanya mencegah paket baru, AI aktif membersihkan repositori lama yang sudah telanjur gemuk.
- Jalankan atau manfaatkan tool
nobloat scandannobloat purge <package>. - Ganti pemanggilan library bloat dengan fungsi bawaan modern, jalankan test suite untuk memvalidasi zero-breakage, lalu hapus dari manifest.
Pillar 2: The Transitive Blast-Radius Rule
Jangan menilai paket hanya dari ukurannya yang tampak kecil.
- Sebelum merekomendasikan sebuah pustaka, periksa transitive dependencies.
- Jika sebuah pustaka untuk tugas sepele menarik $>3$ sub-dependencies pihak ketiga, pustaka tersebut wajib ditolak dan digantikan dengan solusi native mandiri.
Pillar 3: Diff-Only Output Protocol (Token & Context Limit Shield)
Untuk menghemat token context window dan kuota pengguna:
- DILARANG KERAS memuntahkan ulang seluruh file kode jika perubahan $\le 30$ baris.
- Gunakan tool penyuntingan targeted (
replace_file_content) atau tampilkan cuplikan unified diff ringkas. - Hindari basa-basi pembuka dan penutup ("Tentu, saya akan membantu...", "Semoga bermanfaat!").
Pillar 4: Tech-Stack Reality Check (Anti-Framework Bloat)
- Bertindak sebagai Pragmatic Lead Engineer: jika kebutuhan hanya sistem internal dengan pengguna terbatas, tolak arsitektur multi-service terdistribusi yang rumit.
- Prioritaskan kesederhanaan operasional: SQLite dengan mode WAL (Write-Ahead Logging) atau single-node PostgreSQL jauh lebih efisien daripada tumpukan 10 microservices cloud.
Pillar 5: Runtime Compatibility Matrix (Zero-Breakage Guard)
Sebelum menerapkan API native modern, selalu verifikasi target runtime proyek:
- Node 18+:
fetch(),crypto.randomUUID()aman digunakan. - Node 20+:
process.loadEnvFile()dan flag--env-filetersedia. - Node 22+:
fs.promises.glob()tersedia. - Python 3.10+: Pattern matching (
match-case),dataclassesbawaan stabil.
4. Native Standard Library Replacement Matrix
A. JavaScript / TypeScript & Node.js
| Jangan Install (Bloat) | Gunakan Native / Standard Library | Contoh Kode Lean |
uuid | crypto.randomUUID() | const id = crypto.randomUUID(); |
axios / node-fetch | Native fetch() + AbortSignal.timeout() | const res = await fetch(url, { signal: AbortSignal.timeout(5000) }); |
lodash.get / lodash.set | Optional Chaining & Nullish Coalescing | const val = obj?.user?.profile?.name ?? "Guest"; |
lodash.cloneDeep | structuredClone() | const copy = structuredClone(original); |
lodash.flatten | Array.prototype.flat() | const flat = arr.flat(Infinity); |
rimraf | fs.promises.rm | await fs.promises.rm(dir, { recursive: true, force: true }); |
mkdirp | fs.promises.mkdir | await fs.promises.mkdir(dir, { recursive: true }); |
dotenv (Node 20+) | process.loadEnvFile() atau CLI --env-file | process.loadEnvFile(); |
chalk / colorette | ANSI escape codes atau util.styleText | \x1b[32m%s\x1b[0m |
moment / dayjs | Intl.DateTimeFormat | new Intl.DateTimeFormat('id-ID').format(new Date()); |
B. Python (Python 3.10+)
| Jangan Install (Bloat) | Gunakan Standard Library | Keuntungan | |
requests (mikro/webhook) | urllib.request + json | Zero-dependency, langsung jalan tanpa pip install. | |
pydantic (internal container) | dataclasses.dataclass / NamedTuple | Zero runtime parsing overhead, memory footprint minimal. | |
tabulate (tabel CLI) | Format strings / f-strings | `print(f"{name:<20} | {status:>10}")` |
python-dotenv | os.environ + parser sederhana | Menghemat ketergantungan external environment. | |
pytest (unit test mikro) | unittest & unittest.mock | Standard library bawaan Python. | |
pandas (hanya parse CSV kecil) | csv.DictReader | Menghemat download binary 50MB+ dan kompilasi C. |
5. Tool Pendamping: min-tok CLI (Zero-Bloat Engine)
AI dan pengguna dapat memanfaatkan engine zero-dependency min-tok yang berlokasi di C:\Users\MSI\.gemini\antigravity\scratch\nobloat atau dari repositori min-tok:
node bin/min-tok.js scan [dir] [--strict]: Pindaipackage.json,requirements.txt, danpyproject.tomluntuk menemukan bloat libraries.node bin/min-tok.js audit [dir] [--strict]: Audit rasio over-engineering, tingkat abstraksi, dan densitas dependensi.node bin/min-tok.js purge <package> [dir]: Refaktor kode otomatis mengganti pustaka bloat ke native standar.node bin/min-tok.js diff <file1> <file2>: Buat unified diff ringkas dan hemat token.
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.

