---
name: decision-orchestrator-skill
description: Use when the user wants to run a Router→Advisors→CEO decision workflow with minimal manual orchestration, generate structured run artifacts, or prepare OpenClaw-compatible payloads.
---

# Decision Orchestrator Skill

## Overview
This skill reduces multi-file manual orchestration into a single run command. It creates one run package with Router/Agents/CEO artifacts and an OpenClaw handoff payload.

## When To Use
- User asks to run one decision end-to-end with a decision-orchestrator architecture.
- User wants fewer manual prompt-file calls.
- User wants OpenClaw-compatible input/output for workflow automation.

## Workflow
1. Prepare input JSON (see `references/openclaw-input.example.json`).
2. Run `scripts/run_multi_advisor.py`.
3. Use generated artifacts under the run folder.
4. Optionally pass `07-one-shot-prompt.md` to one model call for compact execution.

## Command
```bash
python3 scripts/run_multi_advisor.py \
  --input references/openclaw-input.example.json \
  --runs-root ./runs \
  --owner "Decision Owner"
```

## Artifacts
Per run, this skill creates:
- `01-decision-input.md`
- `02-router-output.yaml`
- `03-agent-outputs.yaml`
- `04-ceo-final.yaml`
- `05-decision-log.md`
- `06-review-summary.md`
- `07-one-shot-prompt.md`
- `openclaw-payload.json`

## UI Display Profiles
- Default profile in `agents/openai.yaml`
- Alternate profile in `agents/openai.chatgpt.yaml`

These profile files control UI metadata only, not Git commit identity.