English | 简体中文

What is this?

Thought Refiner is an open-source Codex Skill for turning an unclear idea, plan, design, or decision into a complete and actionable direction.

Instead of jumping straight into execution, it maps the problem as a decision tree. It asks only the questions that can be answered now, explains each choice in plain language, recommends a default, and waits for confirmation before moving forward.

Why use it?

Capability What it protects
🌳 Dependency-aware questions Later decisions are not asked before their prerequisites are settled
🔎 Fact finding first The agent checks discoverable facts instead of turning research into user homework
🧭 Clear recommendations Every meaningful choice comes with an opinionated default and a short reason
💬 Plain-language rounds Questions stay compact, readable, and aligned with the user's language
🧩 Hidden-assumption checks Unspoken constraints and missing branches become visible before execution
✅ Confirmation gate The final direction is summarized and confirmed before any follow-up work begins

How it works

flowchart LR
    A["Idea, plan, or decision"] --> B["Check discoverable facts"]
    B --> C["Build the decision tree"]
    C --> D["Ask the current frontier"]
    D --> E["User answers"]
    E --> F{"Open branches remain?"}
    F -- Yes --> D
    F -- No --> G["Summarize and confirm"]

The current frontier is the set of decisions whose prerequisites are already settled. Questions that depend on an unanswered decision wait for a later round.

Quick start

Install with the Skill installer

Ask Codex to install this Skill from the repository path:

Use $skill-installer to install:
https://github.com/chenjieliefu/thought-refiner/tree/main/thought-refiner

Install manually

Copy the thought-refiner folder into your personal Codex skills directory:

$HOME/.agents/skills/thought-refiner

Codex normally detects new skills automatically. Restart Codex if it does not appear.

Invoke

Thought Refiner uses explicit invocation, so mention it directly:

Use $thought-refiner to help me decide whether this product idea is worth building.
Keep the questions simple and recommend a default for every choice.

You can also use it for a plan or design:

Use $thought-refiner to stress-test my launch plan before I start execution.

Good fits

  • Product and startup ideas
  • Project or launch plans
  • Product and technical designs
  • Important personal or team decisions
  • Any proposal that still feels under-specified

Repository structure

.
├── README.md
├── README.zh-CN.md
├── LICENSE
└── thought-refiner/
    ├── SKILL.md
    └── agents/
        └── openai.yaml

License

Released under the MIT License.