Coding Standards

An opinionated, model-invoked skill. Creates CODING_STANDARDS.md at the project root by assessing the project, recommending a baseline of MUST / SHOULD / OPTIONAL principles from a curated catalog, then grilling the user to customize every applicable section.

The resulting document is timeless — rules-only, no file paths, no architecture maps, nothing that drifts when the codebase evolves.

Leading words (compact concepts the model already knows — anchor the skill's behavior):

• OCP — add behaviour by creating new files, not modifying existing ones
• Deep Module — simple interface, substantial hidden implementation
• Fail Fast — detect and report errors at the earliest moment
• CQS — a method is either a command (void) or a query (returns data), never both
• YAGNI — don't add abstraction until the code calls for it
• POLA — principle of least astonishment; code should not surprise

Target audience for the document: developers coding with AI agents. The standards are written to reduce ambiguity, provide clear patterns, and give agents a reliable behavioural contract.

Steps

1. ASSESS — understand the project