From wicked-garden
Multi-model AI collaboration: discover installed LLM CLIs and orchestrate council sessions, cross-model reviews, and diverse perspective gathering. Detects codex, copilot, gemini, opencode, and pi CLIs at runtime via PATH discovery. Decisions stored in wicked-garden:mem. Transcripts persisted via jam scripts. Use when: - Running multi-model analysis or design review - Getting diverse AI perspectives on a decision - Council sessions with multiple AI models - Second opinion from a different AI - Multi-model code review or architecture critique - "multi-model", "council", "cross-ai", "diverse perspectives", "second opinion"
npx claudepluginhub mikeparcewski/wicked-garden --plugin wicked-gardenThis skill uses the workspace's default tool permissions.
Orchestrate multi-model AI collaboration using external LLM CLIs.
Provides Ktor server patterns for routing DSL, plugins (auth, CORS, serialization), Koin DI, WebSockets, services, and testApplication testing.
Conducts multi-source web research with firecrawl and exa MCPs: searches, scrapes pages, synthesizes cited reports. For deep dives, competitive analysis, tech evaluations, or due diligence.
Provides demand forecasting, safety stock optimization, replenishment planning, and promotional lift estimation for multi-location retailers managing 300-800 SKUs.
Orchestrate multi-model AI collaboration using external LLM CLIs. Each council member is a different model provider for genuine perspective diversity.
The multi-model system uses external LLM CLIs discovered at runtime:
which codex copilot gemini opencode pi detects installed CLIs# Council mode — dispatches to external CLIs in parallel
/jam:council "Should we use JWT or sessions for auth?"
# Quick jam — single model, 4 personas, fast
/jam:quick "How should we improve the visual design?"
# Full brainstorm — single model, 4-6 personas, 2-3 rounds
/jam:brainstorm "Architecture for the notification system"
| CLI | Install | Model |
|---|---|---|
codex | brew install codex | OpenAI Codex |
copilot | brew install copilot-cli | GitHub Copilot |
gemini | npm i -g @google/gemini-cli | Google Gemini |
opencode | brew install opencode | Configurable |
pi | brew install pi-mono | Pi AI |
Claude always participates as a council member alongside the external CLIs.
┌─────────────────────────────────────────────────────┐
│ /jam:council Command │
│ - Parses topic, options, criteria │
│ - Dispatches to council agent │
├─────────────────────────────────────────────────────┤
│ Council Agent (agents/jam/council.md) │
│ - Detects CLIs via `which` │
│ - Builds question scaffold (4 fixed questions) │
│ - Pipes scaffold to each CLI in parallel │
│ - Claude answers the same scaffold independently │
├─────────────────────────────────────────────────────┤
│ External CLI Dispatch │
│ - cat scaffold.md | codex exec "..." │
│ - cat scaffold.md | gemini "..." │
│ - cat scaffold.md | opencode run "..." │
│ - cat scaffold.md | pi exec "..." │
├─────────────────────────────────────────────────────┤
│ Synthesis (3-stage) │
│ - Stage 1: Raw responses per model │
│ - Stage 2: Synthesis matrix + risk convergence │
│ - Stage 3: Verdict (consensus or fault lines) │
└─────────────────────────────────────────────────────┘
After gathering perspectives from different models, synthesize using:
| Signal | Meaning | Action |
|---|---|---|
| Consensus (2+ models agree) | High confidence issue | Address immediately |
| Unique insight | One model caught it | Evaluate carefully |
| Disagreement | Genuine tradeoff | Human decides |
| Silence | No model flagged it | Lower priority |
Council responses are persisted as transcript entries:
persona_type: councilentry_type: synthesis/jam:transcript and /jam:thinkingStore decisions with full attribution:
/wicked-garden:mem:store "Auth: JWT with 15min/7day expiry.
Consensus: Claude, Gemini, Codex (idempotency critical).
Unique: Gemini flagged session store scaling concern.
Dissent: none." --type decision --tags auth,multi-model-review
| Situation | Recommendation |
|---|---|
| Architecture decisions | Yes — high impact, catch blind spots |
| Security review | Yes — different models flag different risks |
| Important PRs | Yes — diverse review perspectives |
| Visual/UX design | Yes — different aesthetic sensibilities |
| Quick bug fix | No — overhead not worth it |
| Routine code | No — single AI sufficient |
If no external CLIs are detected, council refuses and suggests
/jam:brainstorm (single-model, multi-persona) as an alternative.
With only 1 CLI, it runs as "brainstorm with external guest" with a warning.
Orchestration:
CLI Providers:
Quality: