# Meta Cognition

> Meta-Cognition

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

---

# Meta-Cognition

Audit your own reasoning: notice when shortcuts, mood, or identity are driving conclusions. Separate genuine knowledge from fluent guessing, and pre-commit to what evidence would revise you.

## When to Use
- High-stakes answers, time pressure, strong preferences, or when conclusions arrived suspiciously fast or feel “too neat.”

## Workflow
1. Before answering, scan for likely biases: anchoring on first numbers, availability from recent events, confirmation seeking, sunk cost, overconfidence (Dunning-Kruger at the edges).
2. State a calibrated confidence level with a because-clause (e.g., “~70%—two independent signals agree; one key fact is unverified”).
3. Pre-commit (even mentally) to decision criteria before listing options, when possible.
4. Ask: am I reasoning from evidence and structure, or rationalizing a preferred outcome?
5. Name what would change your mind (specific observations, measurements, or expert inputs).
6. Explicitly flag true uncertainty vs. false confidence; prefer ranges and conditions over brittle absolutes.

## Examples
**Example 1:** Quick architecture pick: note anchoring on a familiar stack; list one disconfirming experiment (spike, load test) before debating further.

**Example 2:** Debate stance: “80% confident—would drop to 40% if study Y replicates under preregistered protocol.”