decide

name	decide
description	Run a decision through the 12-lens framework
trigger	/decide

Decision Framework

Run any decision through the evaluation framework from system/04_decision_frameworks.md. Produces a structured recommendation.

Usage

/decide [decision or opportunity to evaluate]

Process

1. Parse the decision — What's being decided? What are the options?
2. Run relevant lenses — Not all 12 apply to every decision. Select the 4-6 most relevant:
   • Leverage — disproportionate output?
   • Alignment — advances an active priority?
   • Reversibility — can it be undone?
   • Time Sensitivity — does waiting cost anything?
   • Revenue Impact — creates/protects/accelerates revenue?
   • Asymmetry — large upside, bounded downside?
   • Dependency — unblocks other work?
   • Cost — money, time, attention?
   • Compounding — gets better over time?
   • Simplicity — simpler path available?
   • Confidence — how certain are the inputs?
   • Opportunity Cost — what are we NOT doing?
3. Check against principal's decision framework — system/14_decision_framework.md
   • Does it pass the pattern filter?
   • Where does it sit in the priority stack?
   • Does it pass the asymmetry test?
4. Check authority level — system/12_authority_matrix.md
   • Is this Level 1 (just do it), Level 2 (recommend), or Level 3 (human only)?

Output Format

DECISION: [What we're deciding]
OPTIONS: [A vs B vs C]
RECOMMENDATION: [What to do and why]
KEY LENSES: [3-4 lenses that drove the recommendation]
TRADE-OFF: [What we're giving up]
CONFIDENCE: [High / Medium / Low — and why]
AUTHORITY: [Level 1/2/3 — can the AIgent act or needs approval]
NEXT ACTION: [Specific immediate step]

Rules

• Be direct. Recommend, don't waffle.
• If confidence is low, structure it as an experiment, not a commitment.
• If it's Level 3, present the analysis but explicitly say "this is your call."
• Always name the trade-off. Every yes is a no to something else.