Skip to content

chrismaz11/devvit-expert-skill

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

devvit-expert-skill

A standalone LLM skill for Reddit Devvit developers.

This repo packages a Devvit-focused skill that pushes models toward the current Devvit Web workflow instead of mixed legacy examples. The goal is simple: improve code accuracy for Reddit Devvit apps by making the model classify the repo correctly, follow devvit.json, respect the /api versus /internal split, and keep old @devvit/public-api patterns in maintenance-only mode.

What this skill is

  • A reusable skill folder you can drop into Codex or any comparable LLM/developer workflow that supports local skills, system prompts, or reusable instruction packs
  • A Devvit Web-first guide for building, debugging, migrating, and publishing Reddit apps
  • A compact main skill file plus deeper references for architecture, configuration, patterns, and migration

What it optimizes for

  • Correct repo classification: Devvit Web versus legacy public-api
  • Config-first implementation with devvit.json as the source of truth
  • Clear separation between client UX routes and Reddit callback routes
  • Safer handling of Redis, settings, secrets, triggers, scheduler tasks, and external HTTP
  • Fewer hallucinated APIs and fewer mixed-generation Devvit examples

Repo layout

SKILL.md
references/
  api.md
  patterns.md
agents/
  openai.yaml
  • SKILL.md: primary instructions and trigger metadata
  • references/api.md: config, runtime, endpoint, settings, Redis, and docs guidance
  • references/patterns.md: implementation order, route taxonomy, Redis usage, and migration tactics
  • agents/openai.yaml: UI metadata for environments that surface skills in a picker

Installation

Codex

Copy this repo into your Codex skills directory as devvit-expert:

~/.codex/skills/devvit-expert/
  SKILL.md
  references/
  agents/

Other LLM environments

If your editor, agent runner, or chat workflow supports local skills, custom instruction folders, or reusable prompt packs, copy this repo into the equivalent location as devvit-expert.

Manual use

If your environment does not support skill folders, load SKILL.md as a system prompt or reusable instruction file and read the reference files on demand.

Skill stance

This repo intentionally prefers:

  • modern Devvit Web
  • devvit.json
  • client/server split
  • /api/* for app UX
  • /internal/* for menu actions, forms, triggers, and scheduler callbacks

This repo intentionally de-emphasizes:

  • teaching useWebView plus postMessage as the default for new work
  • treating legacy @devvit/public-api examples as the primary source of truth
  • recommending deprecated or older patterns for fresh apps

Legacy guidance is still included, but only for maintaining or migrating existing repos.

Typical use cases

  • “Build a new Devvit app feature”
  • “Migrate this repo from older Devvit patterns”
  • “Why is my menu action not firing?”
  • “What belongs in devvit.json versus server code?”
  • “How should I structure routes, settings, Redis keys, and scheduler tasks?”
  • “Check whether this app is publish-ready”

Notes

  • This is a standalone Devvit developer skill with no product-specific workflow.
  • The source of truth in this repo is the text-based skill content, not a packaged binary artifact.

Contributing

Issues and PRs are welcome if they improve Devvit accuracy, align the guidance with current Reddit platform patterns, or tighten the skill’s default workflow.

About

A Claude / Codex skill that makes any LLM an expert Devvit developer — covers useWebView, Redis, scheduler jobs, Reddit API, and message passing patterns

Topics

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors