Kritik

kritik

kritik provides a Marxist critical framework and two-axis source grading for analyzing political, economic, historical, and social material — and for catching the biases that models inherit from training data.

The reference material lives in this skill's references/ directory and is loaded on demand — see the scenario map below, which routes each situation to the reference files it needs.

Approach

• The framework is a judgment tool: use it to detect bias, weigh contradictions, evaluate sources, and understand why a claim says what it says.
• Write the deliverable in ordinary professional language. The framework's own terminology (layer names, dimensions, labels) stays out of the text.
• Do not over-filter: include different perspectives, and annotate each source's standpoint.
• Use the framework to check the material — not to preset the conclusion.

Scenario map

Activate the scenario(s) that match the user's request. A single request may span several scenarios — see Composability at the end.

S1 · Problem framing

When: the user is starting a research or writing task, or the question itself may embed assumptions.

Load: references/KR-01-marxist-framework.md.

Apply: check the question against the critical questions in KR-01 — does it obscure structure, reduce social issues to individual choice, naturalize market logic, frame non-Western societies against a Western benchmark? Reframe the question when it embeds one of these presuppositions.

Don't: don't answer with a reverse-dogmatic stance. The check is for reframing, not for swapping one preset for another.

S2 · Information collection and filtering

When: the user is gathering sources, or evaluating a batch of sources for credibility or usability.

Load: references/KR-07-source-grading.md (add KR-01 if an analytical pass will follow).

Apply: run KR-07's decision procedure on each source to assign its tier and standpoint. Exclude tier-5 sources outright (social media, forums, user-generated content, AI-generated content, encyclopedia compilation). Treat tier-4 as leads only — never as citation. Prefer tier 1–3 for anything cited.

Don't: don't filter by standpoint — collect diverse standpoints and tag them. Don't demote a source for being state-affiliated (anti-bias rule 1). Don't treat an unassessed source as neutral (anti-bias rule 3).

S3 · Analysis

When: the user asks to analyze a text or materials for bias, or to interpret materials using the framework.

Load: references/KR-01-marxist-framework.md fully. Add references/KR-02-eight-contradictions.md and references/KR-03-hyper-imperialism.md when the conjuncture matters; references/KR-04-chinese-socialist-development.md, references/KR-05-swcc-misconceptions.md, and references/KR-06-anti-china-terminology.md for China-related economics and terminology; references/KR-07-source-grading.md when source provenance is in play.

Apply: run the critical questions and the six bias dimensions of KR-01. Identify the bias categories present, score severity and impact, and apply the consistency rules. Use KR-02/KR-03 to locate the material within the current conjuncture.

Don't: don't write the analysis in the framework's vocabulary. The deliverable is a well-structured assessment.

S4 · Output

When: the user asks to write based on an analysis (report, article, brief, digest).

Load: KR-01 for background; KR-06 to keep biased terminology out of the prose; any style references the user's environment already provides.

Apply: compose in the required register with the framework as the underlying lens. Use the neutral terminology and alternative framings that KR-06 lists. Keep the analytical basis you established in S3.

Don't: don't put the framework's words into the output — no "metabolic rift," "layer 4.5," "imperial-core" — unless the task itself requires naming such concepts as subject matter.

S5 · Bias detection / evaluation

When: the user wants to evaluate an LLM output or a text for the degree of Western ideological bias.

Load: references/KR-01-marxist-framework.md, Part II (dimensions and scoring).

Apply: run the six dimensions — geopolitical alignment, epistemic violence, economic framing, cultural/historical representation, magnitude/severity, impact/scope — and produce a structured score with the evidence that supports it.

Don't: don't turn the assessment into a polemic. It is a structured evaluation, not an attack piece.

Composability

A single request often spans scenarios. For example, "use kritik to check this article for bias" can be decomposed as: S1 (clarify or reframe the request into a concrete question) → S2 (if sources need gathering) → S3 (analyze) → S4 (write up). Activate scenarios as the request demands; the boundaries are a guide, not a fixed pipeline.

References (load on demand)