Help us improve
Share bugs, ideas, or general feedback.
From flow
Evaluates decisions via stance rotation (neutral, advocate, critic perspectives), synthesizes confidence-rated recommendation with next steps. For architectural choices, tech options, build-vs-buy, tradeoffs.
npx claudepluginhub cofin/flow --plugin flowHow this skill is triggered — by the user, by Claude, or both
Slash command
/flow:consensusThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Structured decision evaluation through stance rotation — analyze from advocate, critic, and neutral perspectives, then synthesize into a confidence-rated recommendation with concrete next steps.
Assembles stakeholder perspectives to deliberate complex architectural, technology, or design decisions before brainstorming, surfacing tensions and convergences without forcing choices.
Provides advocate, critic, and neutral stance prompts for multi-perspective analysis of tradeoffs, risks, decisions, and blind spots in code development.
Brainstorms complex architectural decisions by launching parallel agents from diverse perspectives (pragmatist, security, performance) and synthesizes unified optimal solutions.
Share bugs, ideas, or general feedback.
Structured decision evaluation through stance rotation — analyze from advocate, critic, and neutral perspectives, then synthesize into a confidence-rated recommendation with concrete next steps.
| Decision Scope | Mode | Reason |
|---|---|---|
| Bounded, reversible | Sequential (default) | All perspectives in one pass — fast |
| Multi-month or irreversible | Subagent | Three isolated subagents prevent cross-contamination |
| Perspectives suspiciously aligned | Escalate to Subagent | Lack of genuine disagreement signals contamination |
Use subagent mode when: the decision impacts more than 3 months of work, multiple teams are affected, or sequential perspectives align too easily (suspiciously low disagreement likely signals contamination — isolated subagents are required to get genuine divergence).
See references/consensus-strategy.md for full escalation criteria.
Rotate through three perspectives (see references/stance-rotation.md for detailed prompts):
In subagent mode, dispatch three isolated subagents (one per stance) with identical context. Subagents must NOT see each other's output.
Weigh all three perspectives and produce a recommendation:
Before delivering the synthesis, verify:
Decision: "Should we migrate from REST to GraphQL?"
| Perspective | Key Finding |
|---|---|
| Neutral | Current REST API has 47 endpoints; clients use ad-hoc field filtering. GraphQL would reduce over-fetching but adds schema maintenance. |
| Advocate | Mobile clients would cut payload size ~60%. Single endpoint simplifies versioning. Strong ecosystem tooling available. |
| Critic | Team has no GraphQL experience — 2-3 month learning curve. Caching is harder. Existing REST clients need migration path. |
Synthesis:
Add guardrails instructions here.