qe-framework

QE Framework

Query Execute Framework for Claude Code

165 skills | 21 agents | Folder-aware context memory | SIVS quality gate

---

📖 Try it in your browser — 5-minute tour

English · Intro → · Reference → 한국어 · Intro → · Reference → 日本語 · Intro → · Reference → 中文 · Intro → · Reference →

Rendered guides — view in any browser, no install needed.

---

/Qplan  →  /Qgs  →  /Qatomic-run  →  /Qcode-run-task
 Plan       Spec     Execute          Verify

---

Why QE?

Everything you do with AI comes down to two things: Query and Execute.

"Why is this bug happening?" — that's a Query. "Fix it." — that's an Execute. "Run the tests", "commit this", "is this architecture right?" — all just Q and E, over and over.

The problem is, raw AI doesn't do either well on its own. Query without context gives shallow answers. Execute without verification gives unchecked results. And the precision of your question determines the quality of the output — but maintaining that precision every time is exhausting.

QE Framework puts structure between Query and Execute.

  You say             QE does                    You get
─────────────────────────────────────────────────────────
  "what to do"   →   Plan → Spec (refine Q)   →  precise query
  "do it"        →   Implement → Verify → S   →  verified result

You only say what you want. How to ask the right question, how to verify the result — the framework handles that.

┌─────────────────────────────────────────────────┐
│                  QE Framework                   │
│                                                 │
│  ┌───────────┐  ┌───────────┐  ┌─────────────┐ │
│  │  Context   │  │   SIVS    │  │   165+      │ │
│  │  Memory    │  │   Loop    │  │   Skills    │ │
│  │  Manager   │  │   Engine  │  │   Library   │ │
│  └─────┬─────┘  └─────┬─────┘  └──────┬──────┘ │
│        │              │               │         │
│        └──────────────┼───────────────┘         │
│                       │                         │
│              ┌────────┴────────┐                │
│              │   Claude Code   │                │
│              └─────────────────┘                │
└─────────────────────────────────────────────────┘

Problem	QE Solution
Context window fills with irrelevant info	Folder-aware context — loads only what matches your working directory
No structured workflow for complex tasks	PSE Chain — Plan → Spec → Execute → Verify pipeline
Quality depends on prompt quality	SIVS Loop — automated Spec → Implement → Verify → Supervise gate
No model routing control	SIVS Config — route each stage to Claude or Codex independently
Token waste on repeated scans	Context Memory — pre-analyzed project knowledge, auto-refreshed

---

Install

QE Framework is a Claude Code plugin. Install it with two commands:

# 1. Register the marketplace
claude plugin marketplace add inho-team/qe-framework

# 2. Install the plugin
claude plugin install qe-framework@inho-team-qe-framework

That's it. All 165 skills, 21 agents, and hooks are active immediately.

SSH error? If installation fails with Host key verification failed, set git to use HTTPS:

git config --global url."https://github.com/".insteadOf "[email protected]:"

Then retry the install command.

Verify installation:

claude plugin list
# Should show: qe-framework@inho-team-qe-framework ✔ enabled

Update:

claude plugin update qe-framework@inho-team-qe-framework

Uninstall:

claude plugin uninstall qe-framework@inho-team-qe-framework

Alternative: Local development mode

If you're developing or contributing to QE Framework itself:

git clone https://github.com/inho-team/qe-framework.git
cd your-project
claude --plugin-dir /path/to/qe-framework

---

Quick Start

You only need to remember two commands:

/Qinit    # Set up project, choose engine routing (Claude-only or Claude+Codex)
/Qplan    # Start working — the framework guides you through every next step

/Qinit asks how you want to assign engines to each stage (Spec, Implement, Verify, Supervise). Pick Claude-only or mix in Codex — your choice. After that, /Qplan takes over and tells you exactly what to run next.

---