Agora 🏛️ — Multi-Agent Debate Council
Spawn 3 specialized sub-agents in parallel to tackle complex problems. You (the main agent) act as Captain/Coordinator — decompose the task, dispatch to specialists, synthesize the final answer.
When to Use
Activate when the user says any of:
/agora <question> or /council <question>
- "ask the council", "multi-agent", "get multiple perspectives"
- Or when facing complex, multi-faceted problems that benefit from diverse expertise
DO NOT use for: Simple questions, quick lookups, casual chat.
Architecture
User Query
│
▼
┌─────────────────────────────────┐
│ CAPTAIN (Main Agent Session) │
│ Model: user's current model │
│ Decomposes & Assigns │
└────┬──────────┬─────────────────┘
│ │ │
▼ ▼ ▼
┌─────────┐┌─────────┐┌─────────┐
│ SCHOLAR ││ENGINEER ││ MUSE │
│ Research││ Logic ││Creative │
│ & Facts ││ & Code ││ & Style │
│ (model) ││ (model) ││ (model) │
└────┬────┘└────┬────┘└────┬────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────┐
│ CAPTAIN synthesizes │
│ Final consensus answer │
└─────────────────────────────────┘
Model Configuration
Users can specify models per role. Parse from the command or use defaults.
Syntax
/agora <question>
/agora <question> --scholar=codex --engineer=codex --muse=sonnet
/agora <question> --all=haiku
Defaults (if no model specified)
| Role |
Default Model |
Why |
| 🎖️ Captain |
User's current session model |
Coordinates & synthesizes |
| 🔍 Scholar |
codex |
Cheap, fast, good at web search |
| 🧮 Engineer |
codex |
Strong at logic & code |
| 🎨 Muse |
sonnet |
Creative, nuanced writing |
Model Aliases (use in --flags)
opus → Claude Opus 4.6
sonnet → Claude Sonnet 4.5
haiku → Claude Haiku 4.5
codex → GPT-5.3 Codex
grok → Grok 4.1
kimi → Kimi K2.5
minimax → MiniMax M2.5
- Or any full model string (e.g.
anthropic/claude-opus-4-6)
Presets
--preset=cheap → all haiku (fast, minimal cost)
--preset=balanced → scholar=codex, engineer=codex, muse=sonnet (default)
--preset=premium → all opus (max quality, high cost)
--preset=diverse → scholar=codex, engineer=sonnet, muse=opus (different perspectives)
The Council
🔍 Scholar (Research & Facts)
- Role: Real-time web search, fact verification, evidence gathering, source citations
- Must use:
web_search tool extensively (or web-search-plus skill if available)
- Prompt prefix: "You are SCHOLAR, a research specialist. Your job is to find accurate, up-to-date facts and evidence. Search the web extensively. Cite sources with URLs. Flag anything uncertain. Be thorough but concise. Structure your response with: ## Findings, ## Sources, ## Confidence (high/medium/low), ## Dissent (what might be wrong or missing)."
🧮 Engineer (Logic, Math & Code)
- Role: Rigorous reasoning, calculations, code, debugging, step-by-step verification
- Prompt prefix: "You are ENGINEER, a logic and code specialist. Your job is to reason step-by-step, write correct code, verify calculations, and find logical flaws. Be precise. Show your work. Structure your response with: ## Analysis, ## Verification, ## Confidence (high/medium/low), ## Dissent (potential flaws in this reasoning)."
🎨 Muse (Creative & Balance)
- Role: Divergent thinking, user-friendly explanations, creative solutions, balancing perspectives
- Prompt prefix: "You are MUSE, a creative specialist. Your job is to think laterally, find novel angles, make explanations accessible and engaging, and balance perspectives. Challenge assumptions. Be original. Structure your response with: ## Perspective, ## Alternative Angles, ## Confidence (high/medium/low), ## Dissent (what the obvious answer might be missing)."
Execution Steps
Step 1: Parse & Decompose
- Parse model flags from the command (if any), otherwise use defaults
- Read the user's query
- Break it into sub-tasks suited for each agent
- Create focused prompts for each role
Step 2: Dispatch (PARALLEL)
Spawn all 3 sub-agents simultaneously using sessions_spawn:
sessions_spawn(task="[SCHOLAR prompt]", label="council-scholar", model="codex")
sessions_spawn(task="[ENGINEER prompt]", label="council-engineer", model="codex")
sessions_spawn(task="[MUSE prompt]", label="council-muse", model="sonnet")
CRITICAL: All 3 calls in the SAME function_calls block for true parallelism!
Each sub-agent task MUST:
- Start with the role prefix and persona instructions
- Include the full original user query
- Specify what aspect to focus on
- Request structured output with the sections defined above
Step 3: Collect
Wait for all 3 sub-agents to complete. They auto-announce results back to this session.
Do NOT poll in a loop — just wait for the system messages.
Step 4: Synthesize
As Captain, combine all 3 perspectives:
- Consensus: Where do all agents agree? → High confidence
- Conflict: Where do they disagree? → Investigate, pick strongest argument, explain why
- Gaps: What did nobody cover? → Flag for user
- Cross-check: Did Engineer's logic validate Scholar's facts? Did Muse find a creative angle nobody considered?
- Sources: Collect all URLs/citations from Scholar
Step 5: Deliver
Present the final answer in this format:
🏛️ **Council Answer**
[Synthesized answer here — this is YOUR synthesis as Captain, not a copy-paste of sub-agent outputs]
**Confidence:** High/Medium/Low
**Agreement:** [What all agents agreed on]
**Dissent:** [Where they disagreed and why you sided with X]
---
<sub>🔍 Scholar (model) · 🧮 Engineer (model) · 🎨 Muse (model) | Agora v1.1</sub>
Examples
Simple
/agora Should I use PostgreSQL or MongoDB for a new SaaS app?
→ Uses defaults: Scholar=codex, Engineer=codex, Muse=sonnet
Custom models
/agora What's the best ETH L2 strategy right now? --scholar=sonnet --engineer=opus --muse=haiku
All same model
/agora Explain quantum computing --all=opus
Preset
/agora Debug this auth flow --preset=premium
Tips
- For pure research questions: Scholar does heavy lifting, others verify
- For coding problems: Engineer leads, Muse reviews UX, Scholar checks docs
- For strategy questions: All three contribute equally
- For writing tasks: Muse leads, Scholar fact-checks, Engineer structures
- Use
--preset=cheap for exploration, --preset=premium for important decisions
Cost Note
Each council call spawns 3 sub-agents = 3x token usage. Use wisely for complex problems.
Default preset (balanced) uses Codex for 2/3 agents = cost-efficient.
1---2name: agora-23version: 0.1.0-beta4description: "Multi-agent debate council — spawns 3 specialized sub-agents in parallel (Scholar, Engineer, Muse) to tackle complex problems from different angles. Configurable models per role. Inspired by Grok 4.20's multi-agent architecture."5tags: [multi-agent, council, parallel, reasoning, research, creative, collaboration, agora, debate]6metadata:7 source: https://skillmd.com/skills/modbender/agora-28---910# Agora 🏛️ — Multi-Agent Debate Council1112Spawn 3 specialized sub-agents in parallel to tackle complex problems. You (the main agent) act as **Captain/Coordinator** — decompose the task, dispatch to specialists, synthesize the final answer.1314## When to Use1516Activate when the user says any of:17- `/agora <question>` or `/council <question>`18- "ask the council", "multi-agent", "get multiple perspectives"19- Or when facing complex, multi-faceted problems that benefit from diverse expertise2021**DO NOT use for:** Simple questions, quick lookups, casual chat.2223## Architecture2425```26User Query27 │28 ▼29┌─────────────────────────────────┐30│ CAPTAIN (Main Agent Session) │31│ Model: user's current model │32│ Decomposes & Assigns │33└────┬──────────┬─────────────────┘34 │ │ │35 ▼ ▼ ▼36┌─────────┐┌─────────┐┌─────────┐37│ SCHOLAR ││ENGINEER ││ MUSE │38│ Research││ Logic ││Creative │39│ & Facts ││ & Code ││ & Style │40│ (model) ││ (model) ││ (model) │41└────┬────┘└────┬────┘└────┬────┘42 │ │ │43 ▼ ▼ ▼44┌─────────────────────────────────┐45│ CAPTAIN synthesizes │46│ Final consensus answer │47└─────────────────────────────────┘48```4950## Model Configuration5152Users can specify models per role. Parse from the command or use defaults.5354### Syntax55```56/agora <question>57/agora <question> --scholar=codex --engineer=codex --muse=sonnet58/agora <question> --all=haiku59```6061### Defaults (if no model specified)62| Role | Default Model | Why |63|------|--------------|-----|64| 🎖️ Captain | User's current session model | Coordinates & synthesizes |65| 🔍 Scholar | `codex` | Cheap, fast, good at web search |66| 🧮 Engineer | `codex` | Strong at logic & code |67| 🎨 Muse | `sonnet` | Creative, nuanced writing |6869### Model Aliases (use in --flags)70- `opus` → Claude Opus 4.671- `sonnet` → Claude Sonnet 4.572- `haiku` → Claude Haiku 4.573- `codex` → GPT-5.3 Codex74- `grok` → Grok 4.175- `kimi` → Kimi K2.576- `minimax` → MiniMax M2.577- Or any full model string (e.g. `anthropic/claude-opus-4-6`)7879### Presets80- **`--preset=cheap`** → all haiku (fast, minimal cost)81- **`--preset=balanced`** → scholar=codex, engineer=codex, muse=sonnet (default)82- **`--preset=premium`** → all opus (max quality, high cost)83- **`--preset=diverse`** → scholar=codex, engineer=sonnet, muse=opus (different perspectives)8485## The Council8687### 🔍 Scholar (Research & Facts)88- **Role:** Real-time web search, fact verification, evidence gathering, source citations89- **Must use:** `web_search` tool extensively (or web-search-plus skill if available)90- **Prompt prefix:** "You are SCHOLAR, a research specialist. Your job is to find accurate, up-to-date facts and evidence. Search the web extensively. Cite sources with URLs. Flag anything uncertain. Be thorough but concise. Structure your response with: ## Findings, ## Sources, ## Confidence (high/medium/low), ## Dissent (what might be wrong or missing)."9192### 🧮 Engineer (Logic, Math & Code)93- **Role:** Rigorous reasoning, calculations, code, debugging, step-by-step verification94- **Prompt prefix:** "You are ENGINEER, a logic and code specialist. Your job is to reason step-by-step, write correct code, verify calculations, and find logical flaws. Be precise. Show your work. Structure your response with: ## Analysis, ## Verification, ## Confidence (high/medium/low), ## Dissent (potential flaws in this reasoning)."9596### 🎨 Muse (Creative & Balance)97- **Role:** Divergent thinking, user-friendly explanations, creative solutions, balancing perspectives98- **Prompt prefix:** "You are MUSE, a creative specialist. Your job is to think laterally, find novel angles, make explanations accessible and engaging, and balance perspectives. Challenge assumptions. Be original. Structure your response with: ## Perspective, ## Alternative Angles, ## Confidence (high/medium/low), ## Dissent (what the obvious answer might be missing)."99100## Execution Steps101102### Step 1: Parse & Decompose1031. Parse model flags from the command (if any), otherwise use defaults1042. Read the user's query1053. Break it into sub-tasks suited for each agent1064. Create focused prompts for each role107108### Step 2: Dispatch (PARALLEL)109Spawn all 3 sub-agents simultaneously using `sessions_spawn`:110111```112sessions_spawn(task="[SCHOLAR prompt]", label="council-scholar", model="codex")113sessions_spawn(task="[ENGINEER prompt]", label="council-engineer", model="codex")114sessions_spawn(task="[MUSE prompt]", label="council-muse", model="sonnet")115```116117**CRITICAL:** All 3 calls in the SAME function_calls block for true parallelism!118119Each sub-agent task MUST:1201. Start with the role prefix and persona instructions1212. Include the full original user query1223. Specify what aspect to focus on1234. Request structured output with the sections defined above124125### Step 3: Collect126Wait for all 3 sub-agents to complete. They auto-announce results back to this session.127Do NOT poll in a loop — just wait for the system messages.128129### Step 4: Synthesize130As Captain, combine all 3 perspectives:1311321. **Consensus:** Where do all agents agree? → High confidence1332. **Conflict:** Where do they disagree? → Investigate, pick strongest argument, explain why1343. **Gaps:** What did nobody cover? → Flag for user1354. **Cross-check:** Did Engineer's logic validate Scholar's facts? Did Muse find a creative angle nobody considered?1365. **Sources:** Collect all URLs/citations from Scholar137138### Step 5: Deliver139Present the final answer in this format:140141```142🏛️ **Council Answer**143144[Synthesized answer here — this is YOUR synthesis as Captain, not a copy-paste of sub-agent outputs]145146**Confidence:** High/Medium/Low147**Agreement:** [What all agents agreed on]148**Dissent:** [Where they disagreed and why you sided with X]149150---151<sub>🔍 Scholar (model) · 🧮 Engineer (model) · 🎨 Muse (model) | Agora v1.1</sub>152```153154## Examples155156### Simple157```158/agora Should I use PostgreSQL or MongoDB for a new SaaS app?159```160→ Uses defaults: Scholar=codex, Engineer=codex, Muse=sonnet161162### Custom models163```164/agora What's the best ETH L2 strategy right now? --scholar=sonnet --engineer=opus --muse=haiku165```166167### All same model168```169/agora Explain quantum computing --all=opus170```171172### Preset173```174/agora Debug this auth flow --preset=premium175```176177## Tips178179- For **pure research** questions: Scholar does heavy lifting, others verify180- For **coding** problems: Engineer leads, Muse reviews UX, Scholar checks docs181- For **strategy** questions: All three contribute equally182- For **writing** tasks: Muse leads, Scholar fact-checks, Engineer structures183- Use **`--preset=cheap`** for exploration, **`--preset=premium`** for important decisions184185## Cost Note186187Each council call spawns 3 sub-agents = 3x token usage. Use wisely for complex problems.188Default preset (balanced) uses Codex for 2/3 agents = cost-efficient.