# Octo Thinks Decision Matrix

> Turn personal or product choices into transparent weighted decision matrices with criteria, weights, trade-off testing, and clear uncertainty.

- Skill: `mario/octo-thinks-decision-matrix-2` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add mario/octo-thinks-decision-matrix-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mario/octo-thinks-decision-matrix-2/raw
- Safety review: pending
- Works with: any agent that reads SKILL.md (Claude Code, Claude.ai, Cursor, Codex, Windsurf, 60+ more)
- Category: Research & Search
- Author: Mario (https://skillmd.com/u/mario)
- Updated: 2026-10-05
- Page: https://skillmd.com/skills/mario/octo-thinks-decision-matrix-2

---


# Octo Thinks Decision Matrix

Help the user make a choice, not outsource responsibility for it. Keep the
decision explicit, reproducible, and honest about uncertainty.

## Workflow

1. Clarify the decision, horizon, budget or other hard constraints, and what a
   useful outcome looks like. Ask only for information that can materially
   change the result.
2. Use two to five concrete, mutually distinct options. Do not invent product
   models, organizations, prices, or current specifications.
3. Choose three to seven independent criteria. Avoid double-counting the same
   preference under different labels.
4. Elicit non-negative weights that total 100%. If the user cannot provide
   them, offer an explicitly labelled equal-weight starting point.
5. Score every option on the same 1–5 scale. Separate sourced facts, estimates,
   and the user's personal preferences in the explanation.
6. Calculate the weighted result, then test at least one plausible change to
   the most important weight. Do not call a close or sensitive result a clear
   winner.
7. Present the matrix, ranking, key trade-off, uncertainty, and a concrete
   next verification step. The user's final decision remains theirs.

Use the calculation details and response shape in
[references/decision-workflow.md](references/decision-workflow.md). Read it
when scoring a matrix or when the user asks how the method works.

## Boundaries

- Work locally with the information supplied in the conversation or files. Do
  not store, publish, upload, or transmit a user's decision data unless they
  explicitly ask.
- Do not browse or imply that a product price, availability, policy, medical,
  legal, financial, or safety fact is current unless the user asks for research
  and it is verified.
- Treat this as decision support, not professional advice. For high-stakes
  choices, identify the appropriate expert or primary source as the next step.
- When a preset is available, treat it as a reviewable starting point. Preserve
  its version and limitations; adapt criteria and weights to the user rather
  than presenting the preset's default ranking as universal.