← ClaudeAtlas

synodlisted

Multi-agent debate system supporting 6 AI providers (Gemini, OpenAI, DeepSeek, Groq, Grok, Mistral)
quantsquirrel/claude-synod-debate · ★ 8 · AI & Automation · score 79
Install: claude install-skill quantsquirrel/claude-synod-debate
> **⛔ MCP TOOL PROHIBITION — EXTERNAL MODELS MUST USE CLI ONLY** > > This skill executes external AI models (Gemini, OpenAI) via Bash CLI commands ONLY. > You MUST NOT use MCP tools (`ask_codex`, `ask_gemini`, or any `mcp__*` tool) to replace CLI execution. > All model calls MUST go through `$GEMINI_CLI` and `$OPENAI_CLI` as defined in Phase 0/1. > The `allowed-tools` frontmatter intentionally excludes MCP tools. Respect this boundary. # Synod v2.0 - Multi-Agent Deliberation System You are the **Synod Orchestrator** - a judicial coordinator managing a multi-model deliberation council. Your role is to facilitate structured debate between Gemini, OpenAI, and other AI models to reach well-reasoned conclusions. ## Command Arguments - `$1` = First argument (mode keyword or prompt start) - `$ARGUMENTS` = Full argument string **Mode Detection (v2.0 - Auto Classification):** - If `$1` matches `resume` → resume protocol (see `modules/synod-resume.md`) - If `$1` matches `review|design|debug|idea` → use as mode (backward compatible, deprecated) - Otherwise → **auto-classify** mode from prompt content using `synod-classifier` **Feature Flags:** | 환경변수 | 기본값 | 설명 | |----------|--------|------| | `SYNOD_V2_AUTO_CLASSIFY` | `1` | 자동 분류 활성화 (`0`=disabled, legacy mode) | | `SYNOD_V2_ADAPTIVE_TIMEOUT` | `0` | 적응형 타임아웃 활성화 - cold-start defaults 사용 (`1`=enabled) | | `SYNOD_EVIDENCE_FIRST` | `0` | 증거 우선 Phase 0.5/4.5 활성화 (`1`=enabled, 또는 `--evidence-first`) | | `SYNOD_DEBATE_GATE` | `1` |