codex-logical-decision-layer

TL;DR

Use when the task has multiple possible solutions, hidden risk, or high cost of being wrong. Produce a concise decision surface: options, evidence, tradeoffs, recommendation, stop conditions.

Activation

1. Activate on $think, $decide, $options, "think from multiple angles", or "find the best direction".
2. Activate for architecture, refactor, debugging strategy, product workflow, release, and non-trivial UI/BE decisions.
3. Auto-pair with codex-reasoning-rigor when the user asks for deeper reasoning or non-generic output.

Hard Rules

• Do not reveal private chain-of-thought. Output the decision contract only.
• Generate 2-4 plausible options, not an unbounded brainstorm.
• Compare options by evidence, cost, risk, reversibility, and verification.
• Prefer the smallest change that satisfies the goal and can be verified.
• If evidence is weak, state the assumption and the first command/file needed to validate it.
• Stop adding abstractions when the current project evidence does not justify them.

Output Contract

## Decision Surface
| Option | When it wins | Cost | Risk | Evidence needed |
| --- | --- | --- | --- | --- |

## Recommendation
Choose <option> because <repo evidence / constraint>.

## Verification
- <command or observable signal>

## Stop Conditions
- <condition that would change the decision>

Command

python "<SKILLS_ROOT>/codex-logical-decision-layer/scripts/build_decision_matrix.py" --problem "<problem>" --options "A,B,C" --format markdown

Reference Files

• references/decision-contract.md: compact decision matrix rules and anti-token-waste policy.