Skill

craft-squad

Full design-first pipeline with all three AI agents: Claude (architect + UI + integration), Codex (data-layer + bulk fixes), and local LLM (LM Studio, for exploration + review). Power-user mode. Optionally runs Codex adversarial review if codex-plugin-cc is installed.

npx claudepluginhub alexiolan/craft-skills --plugin craft-skills

Tool Access

This skill uses the workspace's default tool permissions.

Preview

Full design-first pipeline with every AI component active. Combines `/craft-duo` (Codex for data layer) with `/craft-local` (LM Studio for review/exploration). Optionally enables a skeptical adversarial review from Codex via `codex-plugin-cc` if installed.

SKILL.md

Similar Skills

craft-duo

Full design-first pipeline with Codex as a co-executor for data-layer tasks. Claude handles architecture and UI; Codex handles types, services, queries, schemas, enums, mappers, and bulk fixes. Use when you want cost relief via Codex without losing Claude's reasoning on UI and integration.

craft-skills

Citadel — Agent Orchestration Harness for Claude Code

Orchestrates coding tasks in Claude Code via /do router with four-tier agents (skill to fleet), campaign persistence across sessions, parallel git worktrees, and 25 skills for autonomous campaigns.

aradotso-trending-skills-37

using-superpowers

178.4k

Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.

3 files

superpowers

Stats

Stars15

Forks0

Last CommitApr 13, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

Craft (squad variant: Claude + Codex + local LLM)

Full design-first pipeline with every AI component active. Combines /craft-duo (Codex for data layer) with /craft-local (LM Studio for review/exploration). Optionally enables a skeptical adversarial review from Codex via codex-plugin-cc if installed.

Profile: claude+codex+llm

Pipeline: Brainstorm → Plan → Develop (hybrid executors + LLM reviews + optional adversarial review) → Browser Test → Report

Requirements

Codex CLI installed: npm i -g @openai/codex

Codex authenticated: codex login or OPENAI_API_KEY

LM Studio running at ${LLM_URL:-http://127.0.0.1:1234}

AGENTS.md at project root (auto-generated)

Optional: codex-plugin-cc installed in Claude Code — enables adversarial review bonus

How this wrapper works

Writes the profile marker, then delegates to the canonical craft pipeline.

Step 1: Write profile marker

echo -n "claude+codex+llm" > .craft-profile

Step 2: Follow the craft pipeline

Read skills/craft/SKILL.md and follow every phase. All profile gates are satisfied: LLM steps run, Codex routing activates, and the develop skill checks for codex-plugin-cc to optionally run an adversarial review.

The user input is: $ARGUMENTS

Expected behavior

All LLM steps run (LM Studio required)

All Codex routing runs (Codex CLI required)

If codex-plugin-cc is installed, develop Step 3.5 invokes /codex:adversarial-review on the diff as a skeptical second-opinion review

If codex-plugin-cc is NOT installed, the adversarial review step is silently skipped — no failure

When to use

Use /craft-squad when:

Stakes are high and you want every layer of review

The feature is complex enough that a skeptical second opinion matters

You have both LM Studio and Codex CLI available

For most day-to-day work, /craft-duo (no LLM) or /craft (no external deps) are usually enough.