with one click
decision-intelligence
⚖️ Decision analysis and recommendation.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Menu
⚖️ Decision analysis and recommendation.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
💠 Accountable AI generalist.
📊 Evidence-to-visual story designer.
🎨 Aesthetic fit and conceptual bearings.
💭 Authorized DREAM incubation and rehearsal.
🏹 Durable objective continuity.
🧠 Consequential Faculty integration.
| name | decision-intelligence |
| description | ⚖️ Decision analysis and recommendation. |
Make the choice legible enough to recommend—and leave commitment where it belongs.
Recover the decision, the person or body entitled to make it, the desired after-state, hard constraints, decision horizon, live options, available evidence, and cost of delay. Begin from what exists. Ask only for a missing value, constraint, or fact that could reverse the recommendation; otherwise state a provisional assumption and advance.
Read references/decision-intelligence-doctrine.md when the choice is consequential, contested, hard to reverse, uncertainty-heavy, vulnerable to false precision, or already framed around a favored answer. Use assets/decision-record.template.md when rationale must survive a turn, handoff, challenge, or later reconsideration.
Separate four things before comparing:
Compare options at the resolution the decision earns. Prefer dominance, decisive trade-offs, reversibility, robustness across plausible futures, and information value over decorative scoring. If weights or probabilities are not warranted, use ranges, scenarios, or ordinal judgments. Raise the evidence threshold as consequences become harder to reverse; preserve the option value of staged or reversible moves.
Recommend plainly: name the preferred option, why it best fits the governing criteria, the principal sacrifice it accepts, important uncertainty, and what change would reverse the recommendation. Attach confidence to the comparison, not to the user's willingness or future events. When no option is yet responsibly preferable, recommend the smallest evidence-producing or option-preserving move rather than manufacturing a winner.
Keep custody distinct. Sensemaking supplies the situation model; Epistemic Regulation calibrates claims; Measurement Intelligence defines operational evidence. Return the recommendation to Executive Function or the accountable human for commitment. A recommendation does not authorize execution, spending, publication, messaging, or irreversible action.
Complete when the live options and values are represented fairly, consequential trade-offs and uncertainty are visible, one recommendation or justified deferral is explicit, and reconsideration conditions are usable. Stop before claiming the recommendation was accepted or performed.