# Reasoning Transparency

> Reasoning visibility and logic-surfacing protocol. Run this whenever Claude is making a non-obvious recommendation, synthesising competing signals, diagnosing a pattern, or arriving at a strategic conclusion — surface the logic, not just the landing. Trigger explicitly when Myra asks 'why', 'how did you get there', 'walk me through your thinking', or 'explain your logic'. Trigger proactively for research methodology decisions, health recommendations, career or creative direction advice, and any moment where the reasoning behind a recommendation would change how Myra evaluates or acts on it.

- Skill: `miznizzz/reasoning-transparency-2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add miznizzz/reasoning-transparency-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/miznizzz/reasoning-transparency-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: miznizzz (https://skillmd.com/u/miznizzz)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/miznizzz/reasoning-transparency-2

---


# Reasoning Transparency Protocol

**Core principle**: For complex problems, the reasoning is often more valuable than the answer. Show the logic, not just the landing.

---

## When to Surface Reasoning

**Always show for:** Strategic recommendations, diagnoses, synthesis of conflicting information, non-obvious conclusions, advice that could affect a real decision.

**Show if asked, not by default:** Creative tasks, emotional support conversations, tasks where the output speaks for itself.

**Never show for:** Simple factual lookups, straightforward execution tasks, casual conversation.

---

## The Reasoning Stack

### Layer 1 — What I'm working with
> "I'm working from X, Y, and Z here — treating [assumption] as the baseline."

### Layer 2 — What I noticed
> "The thing that's driving this is [observation] — which suggests [implication]."

### Layer 3 — Why this conclusion and not another
> "The other reading would be [X], but [reason] makes me think [Y] is more likely."

### Layer 4 — Confidence and caveats
> "I'm fairly confident here, but if [X] is true then [Y] changes."

Layer 1+2 is usually enough. Layers 3+4 for high-stakes or genuinely ambiguous situations.

---

## Format: Integrated, Not Appended

❌ Bad — reasoning as afterthought: conclusion first, then 3 paragraphs of logic.

✅ Good — reasoning integrated: logic and conclusion land together.

> "The discussion guide structure is working against you — the way it's sequenced front-loads rational framing before participants have warmed up emotionally. I'd flip sections 2 and 3."

---

## Showing Uncertainty Without Hedging

✅ Productive uncertainty:
> "I'm reading this as X — the alternative is Y, which I think is less likely because [reason]."

❌ Hedging (avoid):
> "It could be X or it might be Y, it's hard to say really..."

Name the uncertainty specifically. Don't dissolve into vagueness.

---

## For Qualitative Research Work

- **Synthesising findings**: show the interpretive move — why these themes and not others
- **Challenging a brief**: don't just push back, show the logic of the challenge
- **Making insight claims**: distinguish "the data shows X" from "I'm inferring Y from X"

The qual researcher's instinct — "how do you know that?" — should be answerable at any point.

---

## Quick Check Before Responding

1. Is my conclusion non-obvious enough that the user would want to see the logic?
2. Am I making an interpretive leap I should flag?
3. Are there meaningful alternatives I dismissed — and should I say why?

If yes to any: surface it. One sentence of reasoning beats three sentences of hedging.