WayToAI Pangu Distill

Pangu Distill

Extract runnable thinking frameworks from anything

Install it on a coding agent. Distillation yields an activatable Skill, not a summary.

An idea, yourself, or a conversation can all be distilled.

Install View GitHub

Claude Code · Cursor · Codex · OpenClaw · Gemini CLI · Apache-2.0

Pain

Quotes and summaries are not runnable

An agent can recite “persist” and still fail to judge by a framework when it matters.

  1. It recites, it does not judge

    Ask for a trade-off and the answer slides back into slogans.

  2. Quotes without triggers

    You never know when to pick up which line.

  3. Conclusions without bounds

    If failure conditions are unwritten, people pretend it always works.

Not a prettier summary. A Skill the agent can think with.

Demo

Distill “long-termism” and you do not get the word “persist”

One full distillation. Delivery: mental models with triggers, plus honest boundaries.

User      ❯ Distill "long-termism"

Pangu     ❯ Distilling
            ✓ Seven-question intake (idea / decision use / no primary files)
            ✓ Scripted six-stream collection
            ✓ Seven-level extract: origin stories, decisions, failures, tensions
            ✓ 4.5-layer model passed triple verification
            ✓ Independent scoring passed (no self-grading)

            Output: an activatable long-termism Skill
            Models (each with origin story / trigger):
            ① Compound interest only pays those who stay
            ② The quote is not the information
            ③ Persistence ≠ refusing to update the map (heuristic)

            Honest boundary: talking about ten years with no buffer is self-deception; distilling an idea is not becoming a practitioner

Afterward the agent judges by these models, not by “stay the course.”

Capabilities

This is what a runnable framework looks like

Every Skill needs these five layers. Missing the honest boundary means it is unfinished.

  1. Identity card

    First person: who I am, where I started, how others use me.

  2. Mental models

    3–7 models. Each has an origin story, a trigger, reasoning steps, and limits.

  3. Expression DNA

    Syntax, tics, forbidden words. It speaks like the framework.

  4. Decision frame

    Default path, judgment lines, checkable anti-patterns.

  5. Honest boundary

    What it cannot do, information gaps, when not to use it.

A summary stacks conclusions. A framework says when to pick it up and when to put it down.

Install

Install from the repository, then say one line

Follow the GitHub repository. After install, restart or rescan, then say any line below.

  1. Distill "long-termism"
  2. Distill myself
  3. Distill this conversation

The same Skill runs on Claude Code, Cursor, Codex, OpenClaw, and Gemini CLI.

Install from the repo Open repository

Trust

The method can be checked. No scores.

Collection runs scripts. Scoring cannot grade itself.

  1. Primary files first

    Books, transcripts, and chats you provide go first. The web only fills gaps.

  2. Scripted collection

    Default streams: writings, interviews, expression, criticism, decisions, timeline. Failures are recorded.

  3. Honest boundary

    Write what it cannot do and what is missing. Without this layer, it is unfinished.

  4. No self-grading

    Answering and scoring stay apart. You do not write it and then certify it yourself.

This page publishes no scores, no product list, and no prices.

FAQ

Common questions

Who is this for?

People already using a coding agent that can load Skills.

What do I get after install?

An activatable thinking-framework Skill, not a summary or a pile of quotes.

How do I install it?

Open the GitHub repository and follow Quick Install. Commands stay in the repository.

What do I say first?

Distill "long-termism". You can also distill yourself, or this conversation.

How is this different from “summarize this”?

A summary gives excerpts. Distillation gives triggers, reasoning steps, and failure bounds.

If I distill myself or this conversation, will it invent a public persona?

No. It uses the material you provide and this conversation.

Do I have to bring my own sources?

Primary files are better. If the record is thin, that goes into the honest boundary.

Does it cost money?

The repository is Apache-2.0. This page does not promise commercial service.

Which hosts are supported?

Claude Code, Cursor, Codex, OpenClaw, and Gemini CLI. The repository is the source of truth.

Next

A thinking framework should run.

Install from the repository. Then distill “long-termism.”

Install Pangu Distill View GitHub