---
name: meeting-orchestrator
description: |
  Native OpenClaw skill for orchestrating multi-role meetings and single-role analysis using the meeting project roles.
  Use when the user wants to start a meeting, review a proposal with multiple roles, or ask a specific role like real_powell
  to evaluate a topic with human-in-the-loop steps.
user-invocable: true
---

# Meeting Orchestrator

This skill turns the meeting project into a native OpenClaw workflow.

## What it does

- Starts a multi-role meeting with human-in-the-loop checkpoints
- Runs a single-role analysis using a specific role such as `real_powell`
- Uses AI to suggest meeting participants before the user confirms them
- Accepts large topic/background inputs and can summarize them first
- Loads role prompts from the meeting project without invoking its CLI
- Keeps the orchestration logic inside OpenClaw

## Core workflow

1. Load the requested role configs from `/root/projects/meeting/agents/`
2. Let AI suggest a participant list for the topic
3. Show a startup brief and let the user adjust the plan
4. Build a shared scratchpad for the discussion
5. Use a PM router to decide the next role or finish the meeting
6. Allow user feedback between rounds
7. Save the result as a Markdown report

## Recommended user flows

### Start a meeting

Use when you want multiple roles to debate a topic.

Inputs:
- topic
- participant roles or auto-select preference
- optional context
- optional max rounds

### Single-role analysis

Use when you want one role, such as `real_powell`, to evaluate a topic directly.

Inputs:
- role name
- topic
- optional context

## Files

- `references/one-page-reference.md` — compact quick reference
- `references/user-manual.md` — full end-user guide
- `references/frequently-asked-questions.md` — common questions and direct answers
- `references/final-spec.md` — canonical workflow and interaction contract
- `references/quick-start.md` — short usage guide
- `references/entrypoints.md` — how humans should start using the skill
- `references/ingestion.md` — how to feed large background text into the workflow
- `references/verification.md` — manual verification checklist
- `references/meeting.md` — implementation notes and structure
- `references/roles.md` — role catalog and suggested usage
- `references/usage.md` — user-facing usage guidance
- `references/execution.md` — execution contract and hooks
- `scripts/runtime.py` — native workflow data models and execution core
- `scripts/handlers.py` — OpenClaw-facing routing helpers

## Notes

- This skill is designed to avoid direct CLI execution.
- Each speaker turn should be surfaced as a separate standalone message.
- The meeting project remains the source of role definitions.
- Prefer human-in-the-loop prompts before committing to a full run.