LaunchKit · 2026
Back to Skills

roundtable-adaptive

Adaptive multi-model AI roundtable.

0
434 downloads
by @jimmyclanker

Setup & Installation

openclaw skills install @jimmyclanker/roundtable-adaptive

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install roundtable-adaptive

Version History

v2.9.1Feb 26, 2026 - Version note by skill developer:

Trust/consistency patch: aligned runtime requirements with optional providers, clarified persistent-vs-one-shot agent behavior, reduced scanner ambiguity

What This Skill Does

Runs up to 4 AI models in a structured debate with two rounds of cross-critique and formal consensus scoring. Claude is the required base panelist; GPT-5.3 Codex, Grok 4, and Gemini 3.1 Pro are optional additions via Blockrun proxy. Falls back to Claude-only if optional providers are unavailable.

Rather than querying models separately and reconciling answers yourself, the skill runs parallel debate, cross-critique, and synthesis in one automated workflow with a scored consensus output.

When to use it

  • Debating architectural tradeoffs before a major system redesign
  • Red-teaming a security proposal with adversarial AI panelists
  • Choosing between competing product roadmap options with a structured vote
  • Code review where multiple models write then critique each other's implementations
  • Getting multi-model consensus on a high-stakes vendor or hiring decision

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: roundtable --debate Should we migrate our monolith to microservices?

AGENT
  1. 1Creates a Discord thread and checks for duplicate roundtable topics in the last 24 hours
  2. 2Runs a web search to gather current context on monolith vs microservices tradeoffs
  3. 3Spawns a meta-panel of 4 models to design the optimal debate workflow and panel composition
  4. 4Runs 4 panelists in parallel for Round 1, then sends cross-critique prompts to each session in Round 2
  5. 5Spawns a neutral synthesis model to produce a final summary with a calculated consensus percentage
OUTPUT

Discord thread containing each panelist's position, cross-critiques, a consensus score, and a neutral synthesis; full results saved to {workspace}/memory/roundtables/ as JSON

Requirements

Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.

ANTHROPIC_API_KEY environment variableOPENAI_API_KEY environment variable (optional, enables GPT-5.3 Codex panelist)Blockrun wallet funded with USDC on Base (optional, enables Grok 4 and Gemini 3.1 Pro panelists)