Every published module, all in one place.
4 modules
AI Behavior → Code Style Rules
How an AI coding assistant should match a codebase rather than impose its own preferences: mirror existing style, avoid drive-by refactors, keep diffs minimal, comment only where code can't speak, never leave placeholder code, respect the project's stack choices, and ask before adding dependencies.
by @markdowners
Data & AI Features → LLM Integration
How to integrate an LLM API into a production application: never trust model output blindly, streaming UX, retry with backoff on rate limits, token budget management, PII-aware prompt/response logging, fallback behavior on failure, cost monitoring, and model version pinning.
AI Behavior → Prompt Hygiene
How to write prompts that reliably produce the output you want: specific over vague instructions, examples over descriptions, positively stated constraints, stable-first context ordering, explicit output format, iteration over mega-prompts, and testing prompts like code.
Data & AI Features → Prompt Templates
Rules for building reusable LLM prompt templates: variable delimiting, system/user separation, input escaping, versioning, few-shot examples as data, edge-case testing, and output format contracts.