# Demo

> Cofounder demo — fan out curated delegation tasks across Gemini Flash, Claude Opus, Claude Sonnet, and the deterministic ONEX path, then render an ASCII cost comparison chart through native OmniMarket nodes.

- Skill: `omninode-ai/demo` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add omninode-ai/demo`
- Raw SKILL.md: https://api.skillmd.com/api/skills/omninode-ai/demo/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: OmniNode-ai (https://skillmd.com/u/omninode-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/omninode-ai/demo

---


# /onex:demo — Cofounder Delegation Demo

**Announce at start:** "I'm using the demo skill to run the multi-model delegation fan-out."

## Runtime Boundary

`/onex:demo delegation` dispatches through the OmniMarket runtime adapter and the native demo node contracts:

- `node_demo_fanout_orchestrator`
- `node_demo_cost_compute`
- `node_demo_renderer_effect`

The skill shim must not import node handlers directly and must not synthesize a successful response outside the runtime path. Dry-run uses deterministic provider fixtures as fan-out inputs, but the fan-out, cost aggregation, and rendering still run through the native node contracts and local runtime/event-bus boundary.

## Usage

```bash
/onex:demo delegation
/onex:demo delegation --count 2
/onex:demo delegation --dry-run
ONEX_DEMO_DRY_RUN=1 /onex:demo delegation
```

## Curated Tasks

Default tasks when neither `--count` nor `--prompts` is supplied:

1. `Write a pytest test for a function that adds two integers`
2. `Which is cheaper: GPT-4o or Gemini Flash 2.0 for summarization tasks?`
3. `Run the ONEX skill router for ticket triage`

## Outputs

Successful runs return:

- per-task, per-model inference results
- per-model cost breakdown
- `cheapest_llm_model`
- `cheapest_overall_path`
- ASCII chart lines rendered by `node_demo_renderer_effect`
- native node list and run/correlation IDs

## Live Provider Guard

Dry-run is the deterministic, zero-spend proof path. Live mode is explicit and the fan-out node performs provider preflight before any live call. Missing `GEMINI_API_KEY` or missing local `claude` CLI must surface as a structured runtime failure, not an import crash or hidden fallback.