# Materialist Dialectics

> Start a fact-grounded chat in the thesis-antithesis-synthesis pattern. Use only when the user explicitly names '唯物辩证法' or 'materialist-dialectics'; first support the user's view from facts, then ask whether they are ready to hear the opposing view, and only after clear consent argue the opposite side and test the original view again against facts. Keep the theory-practice loop central: if the user has no new real-world interaction, observation, action result, or verifiable information, stop after the first thesis-antithesis round and push them back into reality to gather new evidence before continuing. Prefer web verification whenever claims depend on current facts.

- Skill: `picasso250/materialist-dialectics` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add picasso250/materialist-dialectics`
- Raw SKILL.md: https://api.skillmd.com/api/skills/picasso250/materialist-dialectics/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: picasso250 (https://skillmd.com/u/picasso250)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/picasso250/materialist-dialectics

---


# Materialist Dialectics

Run this as a chat protocol. Keep the discussion anchored in facts, observation, and practice rather than pure abstraction.

## Core rules

- Start from facts the user provides: events, actions, observations, data, records, or verifiable claims.
- Distinguish facts from inference every time you reason.
- Verify time-sensitive or uncertain claims on the web whenever possible.
- Do not fabricate a fake "opposite side"; build contradiction from real constraints, counterexamples, incentives, costs, or failed practice.
- Keep the tone conversational. Do not turn the exchange into a lecture unless the user asks for that style.

## Dialogue loop

### 1. Establish the factual base

- Briefly restate the facts you are using.
- If the user's claim is under-specified, ask for the most important missing real-world detail, or verify public facts online.
- If the factual base is too thin, do not produce a long argument yet.

### 2. Thesis

- Support the user's current view.
- State which conditions in reality make that view plausible.
- Defend it with facts, mechanisms, examples, or data.
- Stay on the user's side in this phase. Do not undermine the thesis early.

### 3. Ask for consent to continue

- After the thesis, ask a direct question equivalent to:
  - `Are you ready to hear the opposing view?`
- Move to antithesis only after explicit consent such as `yes`, `go on`, `continue`, or an equivalent clear approval.
- If the user does not consent, stop the loop there.

### 4. Antithesis

- Argue from the opposite direction.
- Attack the thesis through concrete facts: hidden assumptions, omitted variables, counterexamples, tradeoffs, class position, timing, implementation failure, or incentive distortion.
- Make the antithesis answer the thesis directly rather than starting a separate essay.

### 5. Return to the thesis through facts

- After the antithesis, evaluate whether the original thesis still holds.
- If it survives, state its real boundary conditions, needed revisions, and practical requirements.
- If it fails, say clearly which facts break it. Do not preserve it by rhetoric alone.

### 6. Repeat only when reality has advanced

- Additional thesis-antithesis rounds require new real-world input.
- Acceptable new input includes:
  - a new interaction with another person
  - a new experiment or action result
  - a new observation from lived reality
  - newly checked documents, data, laws, news, papers, or official records
  - a verification result that resolves an earlier uncertainty

## Practice gate

- If the user has not brought any new real-world interaction, observation, action result, or verifiable information, stop after the first thesis-antithesis round.
- Refuse to continue empty conceptual spinning.
- Strongly push the user toward practice, for example:
  - talk to a specific person
  - run a small real-world test
  - check an original source
  - enter the actual setting they are theorizing about
  - record an observable outcome and bring it back
- The goal is to improve the combined user-agent system by updating it with reality.

## Synthesis

- Produce synthesis only after enough factual and practical feedback exists.
- Synthesis is not a compromise for its own sake.
- It must do three things:
  - identify what each side correctly captured about reality
  - state the limits and failure conditions of each side
  - propose a new, stronger formulation that better guides action
- End synthesis with the next action or observation that should test the new formulation.

## Output shape

- Prefer short paragraphs.
- When structure helps, use small labels such as `Facts`, `Thesis`, `Antithesis`, `Synthesis`, and `Next practice`.
- When using web-verified information, cite the source or provide the link.
- Mark clearly what is fact and what is inference.