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
- Activate on
$think, $decide, $options, "think from multiple angles", or "find the best direction".
- Activate for architecture, refactor, debugging strategy, product workflow, release, and non-trivial UI/BE decisions.
- Auto-pair with
codex-reasoning-rigor when the user asks for deeper reasoning or non-generic output.
Hard Rules
- Do not reveal private chain-of-thought. Output the decision contract only.
- Generate 2-4 plausible options, not an unbounded brainstorm.
- Compare options by evidence, cost, risk, reversibility, and verification.
- Prefer the smallest change that satisfies the goal and can be verified.
- If evidence is weak, state the assumption and the first command/file needed to validate it.
- Stop adding abstractions when the current project evidence does not justify them.
Output Contract
## Decision Surface
| Option | When it wins | Cost | Risk | Evidence needed |
| --- | --- | --- | --- | --- |
## Recommendation
Choose <option> because <repo evidence / constraint>.
## Verification
- <command or observable signal>
## Stop Conditions
- <condition that would change the decision>
Command
python "<SKILLS_ROOT>/codex-logical-decision-layer/scripts/build_decision_matrix.py" --problem "<problem>" --options "A,B,C" --format markdown
Reference Files
references/decision-contract.md: compact decision matrix rules and anti-token-waste policy.
1---2name: codex-logical-decision-layer3description: Use when work is ambiguous or has multiple viable paths; compares options, assumptions, evidence, tradeoffs, and the best next action.4---56## TL;DR7Use when the task has multiple possible solutions, hidden risk, or high cost of being wrong. Produce a concise decision surface: options, evidence, tradeoffs, recommendation, stop conditions.89## Activation101. Activate on `$think`, `$decide`, `$options`, "think from multiple angles", or "find the best direction".112. Activate for architecture, refactor, debugging strategy, product workflow, release, and non-trivial UI/BE decisions.123. Auto-pair with `codex-reasoning-rigor` when the user asks for deeper reasoning or non-generic output.1314## Hard Rules15- Do not reveal private chain-of-thought. Output the decision contract only.16- Generate 2-4 plausible options, not an unbounded brainstorm.17- Compare options by evidence, cost, risk, reversibility, and verification.18- Prefer the smallest change that satisfies the goal and can be verified.19- If evidence is weak, state the assumption and the first command/file needed to validate it.20- Stop adding abstractions when the current project evidence does not justify them.2122## Output Contract2324```markdown25## Decision Surface26| Option | When it wins | Cost | Risk | Evidence needed |27| --- | --- | --- | --- | --- |2829## Recommendation30Choose <option> because <repo evidence / constraint>.3132## Verification33- <command or observable signal>3435## Stop Conditions36- <condition that would change the decision>37```3839## Command4041```bash42python "<SKILLS_ROOT>/codex-logical-decision-layer/scripts/build_decision_matrix.py" --problem "<problem>" --options "A,B,C" --format markdown43```4445## Reference Files46- `references/decision-contract.md`: compact decision matrix rules and anti-token-waste policy.
Bang-isme