Skip to content
Conduit

skill

universal-ai-engineer

Cross-platform end-to-end AI/software engineering workflow skill for Claude Code (plugin marketplace, 10 modes).

shinigamijoy/universal-ai-engineercommunitylistedunreviewed
Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/shinigamijoy/universal-ai-engineer itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 17, 2026

universal-ai-engineer

A Claude Code skill that turns ad-hoc AI coding into a disciplined, safety-gated, end-to-end engineering workflow for any software project (web apps, CLIs, libraries, APIs, services, data pipelines, AI/RAG/agent systems) on any OS (Windows / WSL / Linux / macOS).

Distributed as a Claude Code plugin marketplace.

What it does

project setup → environment prep → safe dependency install → architecture check → implement → test → review (Codex or built-in) → fix confirmed issues → retest → ship-check → summary

10 modes: preflight · analyze · start-project · prepare-env · implement · review · refresh-docs · ship-check · self-upgrade · ensure-skills

Highlights:

  • Safety-gated — plans and asks before editing; asks once before any global/system install; never touches .env/secrets; you make the final call.
  • Token-light — progressive disclosure: a lean SKILL.md core + reference files loaded only for the active mode.
  • External-prerequisites check — flags accounts/subscriptions/API keys/services the code needs but the repo lacks, and asks you.
  • Per-stage notifications — a short status at each milestone (plan/env/tests/review/ship/ complete), on the PC by default, or WhatsApp (CallMeBot/Twilio/Meta) if you set it up.
  • draw.io visuals — generates architecture/plan diagrams (.drawio) you can open and edit.
  • Codex-optional review — uses the codex CLI if you've signed in, else the built-in /code-review + /security-review.

Install

/plugin marketplace add shinigamijoy/universal-ai-engineer
/plugin install universal-ai-engineer@universal-ai-engineer-marketplace

(You can also just copy plugins/universal-ai-engineer/skills/universal-ai-engineer/ into your ~/.claude/skills/ directory.)

Usage

/universal-ai-engineer start-project ~/projects/my-app     # new project
/universal-ai-engineer analyze .                           # assess an existing folder (read-only)
/universal-ai-engineer implement add user login            # build a feature / fix a bug
/universal-ai-engineer review current changes
/universal-ai-engineer ship-check                          # before commit/PR

What it runs / permissions (transparency)

  • Edits are gated: it shows a plan and waits for your approval before changing files.
  • Installs: project-local (e.g. .venv, node_modules) auto after showing the command; global/system installs are never silent — it shows the command and asks once.
  • Secrets: never read, stored, logged, or committed; WhatsApp/Codex credentials live only in your environment (env vars / the codex CLI), never in this skill or in chat.
  • Helper scripts (stdlib-only Python, no dependencies):
    • scripts/notify.py — optional WhatsApp / PC status notifier.
    • scripts/visualize.py — generates a .drawio (or HTML) plan/architecture diagram.
  • Per-device setup (notifications, codex login, git) is opt-in and runs at the start of a project (or via preflight); nothing is required for the core workflow.

Versioning

Current: 1.7.0 — see the CHANGELOG section at the bottom of plugins/universal-ai-engineer/skills/universal-ai-engineer/SKILL.md.

License

MIT — see LICENSE.

↑↓ navigate openesc close