Structured Socratic debate for resolving ambiguities, trade-offs, and low-confidence decisions. Produces auditable thesis-antithesis-synthesis records with constitutional principle alignment. [EXPLICIT] Trigger: 'debate', 'ambiguity', 'resolve conflict', 'low confidence', 'Socratic'
"The unexamined decision is not worth implementing."
Formal mechanism for resolving ambiguities that have divergent implementation consequences. Runs a structured thesis → antithesis → counter-evidence → synthesis cycle, examines each option against constitutional principles, eliminates by contradiction, and produces a single answer with full rationale. Records debates in .specify/debates/ for auditability. Use when confidence < 0.95, requirements are ambiguous, or trade-offs have no obvious winner. [EXPLICIT]
.specify/debates/debate-YYYY-MM-DD-topic.md:
# Debate: {topic}
Date: {date}
Trigger: {what caused the debate}
Thesis: {position A}
Antithesis: {position B}
Evidence: {findings}
Constitutional Alignment: {principle mapping}
Synthesis: {final answer}
Confidence: {score}
Integrated Into: {ADR-NNN / plan / spec reference}.specify/debates/| Anti-Pattern | Why It's Bad | Do This Instead |
|---|---|---|
| Skipping the antithesis | Confirmation bias — you only validate your assumption | Always generate the strongest opposing position |
| Debating without evidence | Opinions are not decisions | Ground every position in code, config, or documentation |
| Closing at < 0.95 | Unresolved ambiguity propagates downstream | Keep debating or escalate to user |
| Not recording the debate | Future decisions lose context | Always write to .specify/debates/ |
| Debating trivial choices | Overhead without value | Only debate when implementation consequences diverge |
trade-off-analysis — Weighted decision matrices for architecture choicesscenario-analysis — Multi-scenario comparison with scoringrequirements-engineering — When the ambiguity is in requirements, not solutionsintegrity-chain-validation — Debates may reveal integrity chain gapsExample invocations:
| Scenario | Handling |
|---|---|
| Empty or minimal input | Request clarification before proceeding |
| Conflicting requirements | Flag conflicts explicitly, propose resolution |
| Out-of-scope request | Redirect to appropriate skill or escalate |
Capas del packet, cargables bajo demanda (disciplina ICM: una capa por vez, nunca todas juntas): knowledge/ cuerpo de conocimiento · prompts/ prompts listos · examples/ salida de ejemplo · agents/ subagentes del packet · templates/ plantilla de output.
e8f986b
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.