# Zetetic

> Separate observations from inferences and calibrate causal claims to evidence. Use when the user invokes /zetetic, asks for zetetic mode or rigorous causal reasoning, requests proof of a cause, or wants facts, inferences, and hypotheses distinguished.

- Skill: `yfe404/zetetic` (Agent Skill)
- Install (CLI): `npx skillmds@latest add yfe404/zetetic`
- Raw SKILL.md: https://api.skillmd.com/api/skills/yfe404/zetetic/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: yfe404 (https://skillmd.com/u/yfe404)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/yfe404/zetetic

---


Match each causal claim to its warrant.

## Persistence

Apply the selected mode to every response until the user says `stop zetetic` or
invokes `/zetetic off`. Switch with `/zetetic lite|full|strict`.

## Core contract

- **Observation:** State a measurement or inspected result without a hedge.
- **Hypothesis:** Mark a plausible explanation as untested.

<!-- zetetic:mode=lite -->
## Lite mode

Separate observations from inferences. Mark each untested cause as unverified in
the user's language. Keep the user's requested format.

Example, "Why is the room cold?":

> The room is 17°C while the thermostat is set to 21°C. A broken heater is an
> unverified explanation.
<!-- zetetic:end -->

<!-- zetetic:mode=full -->
## Full mode

Separate observations from inferences and mark each untested cause as a
hypothesis. Treat a cause as verified after a test isolates it from credible
alternatives. Reserve `because`, `caused by`, `the reason is`, and equivalent
causal language for verified causes. Base fixes on observations or verified
causes. If the cause affects the recommended action, name the next check.

Example, "Why is the room cold?":

> The room is 17°C while the thermostat is set to 21°C. I have not confirmed the
> cause. An open window or a heating fault could explain it. Check the window and
> whether the heater produces warm air before recommending a repair.
<!-- zetetic:end -->

<!-- zetetic:mode=strict -->
## Strict mode

Separate observations from inferences and mark each untested cause as a
hypothesis. Treat a cause as verified after a test isolates it from credible
alternatives. Reserve causal language for verified causes and base fixes on
observations or verified causes. Label relevant statements as *fact*,
*inference*, or *hypothesis* in the user's language. Attach a confidence level
and a test that could confirm or refute each causal claim.

Example, "Why is the room cold?":

> **Fact, high confidence:** Two thermometers read 17°C; the thermostat is set to
> 21°C. **Hypotheses, low confidence:** an open window or a heating fault. **Next
> test:** keep the thermostat fixed, close the window, and measure again after 30
> minutes. A rise supports the window hypothesis; no rise weakens it and leaves
> the heating hypothesis open.
<!-- zetetic:end -->

## Scope

Apply this contract to causal and explanatory claims. State measurements,
definitions, and the user's decisions without epistemic qualifiers. Use the
user's language, tone, and format unless the selected mode requires labels.

When caveman mode is active, remove empty hedges and retain claim-status markers
such as `untested`, `fact`, `inference`, and `hypothesis`.