discernment_skill

Overview

This skill convenes internal perspectives before answering questions that carry ethical weight, trade-offs, or potential harm. It prioritizes discernment over a single confident conclusion and surfaces tensions so the user understands the reasoning landscape. Use it whenever a question reasonably admits multiple valid approaches or serious consequences.

How this skill works

On invocation, the skill identifies which internal voices (e.g., Pragmatist, Harm-aware, Long-term thinker) are relevant to the specific question. Each voice speaks once, no rebuttals; silence is allowed. A clerk listens for alignment, real tensions, and possible blocks, then produces an outcome: unity, noted stand-asides, a block, or an escalation with clearly framed tensions.

When to use it

• Life decisions with major consequences (career, relationships, large purchases)
• Ethical dilemmas or questions about potential harm
• Design or architecture choices with multiple valid approaches
• Advice situations where outcomes depend strongly on context (‘it depends’ cases)
• Any request where you might otherwise give a single, confident opinion without showing trade-offs

Best practices

• Choose context-emergent voices rather than a fixed roster for each question
• Limit each voice to one statement and prohibit rebuttals to preserve clarity
• Pause after voices speak; let silence reveal what matters
• As clerk, surface alignment, real tensions, stand-asides, and any blocks explicitly
• When unity isn’t reached, escalate with clear contrasts and a targeted question to the user

Example use cases

• Deciding between hiring approaches where fairness, speed, and cost conflict
• Choosing system architecture (monolith vs microservices) with team and scale trade-offs
• Advising a user weighing a high-risk medical or legal decision where harm is possible
• Resolving product-policy questions that affect user safety and business objectives
• When a user asks for a ‘quick’ answer but stakes suggest careful consideration