Skip to content
Back to skills

Decision Room

ASecurity

Examine a decision or proposition through four voices chosen for the topic. Use for "four people in a room," decision quadrants, competing arguments, or requests to expose weak reasoning and compare options.

  • 5 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 22, 2026
ai-agentsgo

Security analysis

A100/100

Scanned September 22, 2026

npx -y skills add iurysza/agent-skills --skill decision-room --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Decision Room?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Decision Room
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/iurysza-decision-room/badge)](https://www.skillsdirectory.com/skills/iurysza-decision-room)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: decision-room
description: >-
  Examine a decision or proposition through four voices chosen for the topic.
  Use for "four people in a room," decision quadrants, competing arguments,
  or requests to expose weak reasoning and compare options.
metadata:
  category: planning-architecture
---

# Decision room

Show the competing arguments, including tempting ones that fail. Explain what
holds up and what that means for the question. Choose the voices from the topic,
not a fixed list of professions or personality types.

Use the context already provided. State assumptions and proceed unless a missing
fact prevents useful or safe analysis.

## 1. Find the actual question

Separate bundled decisions. Distinguish choosing an action from checking a claim.
State the goal, constraints, and consequences of doing nothing where relevant.

Locate the disagreement. It may concern facts, predictions, values, interests,
or acceptable risk. Different priorities are not automatically reasoning errors.

Separate supplied facts, verified evidence, assumptions, and unknowns. Check
claims that could change the answer when sources are available. Cite the evidence
you use. Never present supplied context as independently verified.

## 2. Choose a map that fits

For a genuine two-sided question, use position and reasoning quality as the axes:

| | Position A | Position B |
|---|---|---|
| Strongest available case | The best-supported argument for A | The best-supported argument for B |
| Weak case worth examining | A tempting argument for A that fails or lacks support | A tempting argument for B that fails or lacks support |

Name A and B after the actual positions. Judge arguments before placing them.
The rows describe particular arguments, not people's intelligence, honesty,
or permanent character.

The strongest available case may still be weak. Do not assume equal support on
both sides or invent a weak argument to fill a cell. Label illustrative examples
as hypothetical. Leave cells empty when necessary.

This matrix is a useful tool, not a rule that every debate must obey. When there
are more than two positions, separate decisions, or compatible choices, use an
option table or decision tree. Keep the argument audit. Include doing nothing,
staged approaches, or selective choices when relevant. Do not assume the middle
position is correct.

## 3. Build the room

Create four hypothetical voices when the material supports them. Use the matrix
cells when they fit. Otherwise, choose voices with different positions, concerns,
or assumptions. Use fewer rather than repeat an argument or invent a caricature.

Give each voice a short label that belongs to this topic. For each, show:

- What they support and what matters to them.
- Their central argument and the evidence or assumption it depends on.
- The strongest objection they need to answer.
- What evidence or changed circumstances would justify reconsidering, and any
  condition they regard as unacceptable.

For a disagreement about values, identify the priority at stake. Do not demand
that a factual finding settle a value judgment.

Use arguments from the supplied discussion first. Distinguish sourced arguments,
your reconstruction, and invented examples. Do not fabricate quotes or attribute
hypothetical arguments to real people or groups. A hypothetical deal-breaker is
not an established constraint. Agreement among the voices is not evidence.

Make weak arguments recognizable, not ridiculous. Explain why they might appeal
to someone. Show the original weakness before offering a repaired version.

## 4. Test the arguments

Use a compact table. Combine it with the voices when that avoids repetition.

| Argument | Evidence or assumption | What holds or fails | What would resolve or repair it? |
|---|---|---|---|

Check the premises, the step from premise to conclusion, and hidden assumptions.
Address the strongest objection, not an easier substitute. Explain the exact
failure in plain language. Naming a fallacy is not enough.

Keep these distinctions clear:

- A contradicted claim differs from one that lacks evidence.
- A conclusion can be true even when the argument for it fails.
- A conclusion can follow from premises that have not been established.
- A value disagreement is not automatically a factual or logical mistake.

For a weak argument, identify what to discard, what narrower claim survives,
and what evidence or change to the argument could repair it. Do not assume every
argument can be repaired. For a stronger argument, state its limits and strongest
counterevidence. Apply the same checks to the option you favor.

Test universal propositions against counterexamples and unclear definitions.
Do not turn a metaphor into a logical law. Confidence, polished language, and
stated willingness to reconsider do not establish correctness. Judge the claim
without guessing the speaker's motives.

## 5. Check the options separately

A bad argument for A does not make A a bad option or establish B. A persuasive
case can still propose something that violates a constraint.

Compare options against the actual goal and priorities. Choose criteria from the
topic rather than reuse a fixed checklist. Distinguish stated priorities from
ones you assumed. Explain consequences, who bears them, and uncertainties that
could change the comparison.

Separate options that work now, need named changes, or violate an established
constraint. Do not count planned safeguards as existing ones. When dismissing an
option, state the specific reason and check whether a different priority or an
overlooked benefit changes the judgment.

Use scores only when requested or when they help explain a choice. Define the
criteria and scale. Keep argument quality separate from option suitability.
Mark unknowns rather than invent precision. Do not average scores by default
or let a total compensate for violating a hard constraint.

## 6. State what survives

End with the arguments to keep, qualify, or discard and the decisive reasons.
State any unresolved disagreement. Unequal evidence does not require a balanced
verdict, and agreement does not need to be manufactured.

For a practical decision, recommend a next step under the stated priorities.
Give the strongest remaining objection and any tempting alternative to avoid.
Name observable conditions that would justify reconsidering. When uncertainty
blocks commitment, propose the smallest useful check or reversible trial, with
a stop condition where relevant. More automation or risk is not inherently better.

For a proposition, explain which parts hold, need qualification, or fail. Offer
a corrected formulation when useful. Do not invent an action decision.

## Writing

Write plainly. Keep tables compact. Skip fictional dialogue, biographies, and
repeated summaries. Explain failures without sneering at the people making them.
Use sentence case and straight quotes. Avoid em dashes, decorative language,
and generic conclusions.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.