<< All versions

Skill v1.0.0

currentLLM-judged scan90/100
openreachtech/hora-skills-ort-core/hoc-charters-coding
──Details
PublishedOctober 21, 2026 at 01:24 PM
Content Hashsha256:f23f5d459a20684a...
Git SHAb211ba9cc372
──Files
Files (1 file, 3.1 KB)
SKILL.md3.1 KBactive
SKILL.md · 55 lines · 3.1 KB

version: "1.0.0" name: hoc-charters-coding description: "ORT coding charter. Write readable, unified code; avoid modification; let the code explain everything."


ORT Coding Charter ⛩️

The coding charter for ORT development. It aims to minimize the reader's burden while maximizing maintainability and AI learning efficiency.

(1) Neither Quirk Nor Quiz

  • Don't write quiz-like code that forces readers to rack their brains, nor unique, unconventional, show-off code.
  • Write simple, unified code that anyone can easily understand.

(2) Writing Once, Reading Unlimited

  • Code is written to be read by others. Hiding information out of laziness only increases the reader's effort.
  • Our job is to write code, not to omit it. Syntactic sugar is the furthest thing from our values.
  • However, widely adopted and established constructs — such as class (sugar over prototype inheritance) or async/await (broadly, sugar over Promises) — are not regarded as syntactic sugar. For constructs whose "sugar-ness" differs between the broad and narrow senses, use only those approved on a whitelist.

(3) Write Code Like Carving on a Slab

  • Once written, code should be modified as little as possible. Overlooking an affected area leads to hard-to-find bugs.
  • If a feature can be added without touching existing code, bugs are confined to the "added code." Learn the Open-Closed Principle and design as if carving into a slab.

(4) Let the Code Explain All

  • Don't write code that relies on implicit understanding or undocumented assumptions. Give the reader all the information needed to understand its purpose and functionality.
  • t() tells us nothing; i18n.t() gives at least a basic sense of what it does.

(5) Write Code You Can Justify

  • The PR author is responsible for justifying their code, able to explain every line when asked "why this logic?".
  • Unacceptable: "I wrote it as told", "I pasted AI output as-is", "it seemed to work", "I reused existing code without understanding it".

(6) Count Spaces as Code

  • Here, "spaces" means half-width spaces and blank lines. Treat them as spaces that contain code, not gaps between code.
  • Spaces clarify structure and reduce the reader's burden. Don't omit blank lines to save keystrokes.

(7) Vertical is Better than Horizontal

  • Prefer vertical code over horizontal. Long lines force horizontal scrolling and disrupt reading flow; the human eye is not accustomed to reading horizontally separated information.
  • Examples: don't pack multiple lines into one; break method chains one by one; put each ternary term on its own line; prefer more records over more DB columns; put each HTML attribute on its own line.

(8) Keep Everything One by One

  • Making the reader extract multiple pieces of information from one piece of code is a trap for oversight. Narrow information down to one to reduce the reader's burden.
  • Examples: one expression per line; don't nest if statements; one responsibility per method; one concept per class; break )) between the two characters; RDB normalization; one purpose per PR.

Version

  • v1: 2025/04/03
All versions