Stakeholder Update Agent

A profile-driven Declarative Agent for M365 Copilot that automates recurring stakeholder updates by aggregating data from multiple sources — Teams meetings, Slack discussions, Azure DevOps work items, and Loop notes — then outputting a voice-calibrated, structured update to a Loop page.

Built for the monthly Foundation Stakeholder Meeting, but designed to be adaptable to any update purpose, cadence, or audience via swappable profiles, templates, and voice profiles.

Architecture

User (M365 Copilot Chat)
  │
  ▼
Declarative Agent
  │
  ├─ instruction.txt        → Profile-aware workflow, voice calibration, few-shot examples
  │
  ├─ profiles/              → JSON configs per meeting type (cadence, audience, data sources)
  │   ├─ foundation-stakeholder.json
  │   ├─ weekly-team-sync.json
  │   └─ example-profile.json
  │
  ├─ templates/             → Output format templates per profile
  │   ├─ foundation-stakeholder.txt
  │   ├─ weekly-team-sync.txt
  │   └─ example-template.txt
  │
  ├─ voice-profiles/        → Writing style guides (tone, structure, language)
  │   ├─ default.txt
  │   └─ example-voice.txt
  │
  └─ plugins/               → MCP data source connectors
      ├─ workiqPlugin.json   → Teams meetings, messages, documents (via WorkIQ)
      ├─ adoFlowPlugin.json  → Work items, sprints, boards (via Azure DevOps)
      ├─ loopPagesPlugin.json→ Raw notes input + formatted output (via Loop)
      └─ slackPlugin.json    → Channel discussions, announcements (via Slack)

How It Works

1. You trigger the agent from M365 Copilot Chat: "Generate my monthly update"
2. Agent loads profile — determines lookback period, data sources, template, and voice
3. Agent reads raw notes from your Loop page (primary grounding context)
4. Agent gathers meeting context via WorkIQ (Teams meetings, messages, docs)
5. Agent pulls work item status via ADO-Flow (Azure DevOps)
6. Agent searches Slack for relevant discussions, decisions, and announcements
7. Agent synthesizes everything using your template and voice profile
8. Agent writes the formatted update to your output Loop page

Prerequisites

• Node.js 18+
• Microsoft 365 Copilot license
• Teams Toolkit VS Code Extension or CLI
• Access to the data sources you want to use (Azure DevOps, Teams, Loop, Slack)

Quick Start

1. Clone and Open

git clone https://github.com/docooley/stakeholder-update-agent.git
cd stakeholder-update-agent
code .

2. Configure Environment

cp env/.env.dev.example env/.env.dev

Edit env/.env.dev and fill in your MCP server URLs and Loop page IDs. See the setup guides for each data source.

3. Customize Your Profile

Edit or copy a profile in appPackage/profiles/:

• Set your ADO organization, project, and team
• Set your Slack channels and search queries
• Set your Loop page IDs
• Choose your template and voice profile

4. Provision

Open in VS Code → Teams Toolkit sidebar → Provision (select dev environment).

Note: Some organizations block CLI-based auth. See Provisioning Guide for VS Code-based provisioning.

5. Test

Open M365 Copilot Chat → find "Stakeholder Update" agent → "Generate my monthly update"

Customization

Create Your Own Profile

1. Copy appPackage/profiles/example-profile.json
2. Rename it (e.g., quarterly-okr-review.json)
3. Set the id, name, cadence, audience, and lookbackDays
4. Configure dataSources — choose which sources to pull from
5. Point to your template and voice profile
6. Fill in sourceConfig with your Slack channels, ADO project, Loop pages, etc.

Create Your Own Template

1. Copy appPackage/templates/example-template.txt
2. Define the sections your audience expects
3. Use placeholder tokens ({Month}, {Workstream Name}, etc.)
4. Reference it in your profile's template field

Calibrate Your Voice

1. Copy appPackage/voice-profiles/example-voice.txt
2. Describe your tone, structure patterns, and language preferences
3. Paste 2-3 excerpts from your own past updates as few-shot examples
4. Reference it in your profile's voiceProfile field

Project Structure

Path	Description
appPackage/declarativeAgent.json	Agent config — name, instructions, plugin references
appPackage/instruction.txt	System prompt — profile-aware workflow, voice, few-shot examples
appPackage/manifest.json	Teams app manifest
appPackage/profiles/	Profile configs (one per meeting type)
appPackage/templates/	Output format templates
appPackage/voice-profiles/	Writing style guides
appPackage/plugins/	MCP data source plugin manifests
docs/	Setup guides for MCP servers and provisioning
env/.env.dev.example	Template environment file
env/.env.dev	Your environment variables (gitignored secrets in .env.dev.user)
teamsapp.yml	Teams Toolkit lifecycle config

Data Sources

Plugin	Source	What It Provides
WorkIQ	Teams meetings, messages, docs	Meeting decisions, action items, shared documents
ADO-Flow	Azure DevOps	Work item status, sprint progress, blockers
Loop Pages	Microsoft Loop	Raw notes (input) + formatted update (output)
Slack	Slack workspace	Channel discussions, announcements, decisions

Each data source is optional — configure only the ones relevant to your workflow.

Contributing

See CONTRIBUTING.md for guidelines on adding profiles, templates, voice profiles, plugins, and documentation.

License

MIT

References

• Declarative Agents for M365 Copilot
• Build Declarative Agents with MCP
• Azure DevOps MCP Server
• Teams Toolkit