cs-context-packet

Imported from davideldersveld/Workplace-AI-Process-Decomposition-Framework (LineOfBusinessSamples/CopilotCoworkSampleSkills/CustomerService/cs-context-packet/SKILL.md). Install upstream with npx skills add davideldersveld/Workplace-AI-Process-Decomposition-Framework --skill cs-context-packet. Copyright stays with the author.

Overview

Assembles a comprehensive context packet by searching across M365 for customer profile data, account history, entitlement and service tier details, prior case history, product or service context, and known issues or active incidents. Produces a Word document that serves as the foundation for classification, severity assessment, and response drafting.

This skill operates in "AI act within policy" mode — it retrieves and assembles approved context but does not classify, prioritize, or recommend actions.

When to Use

• A case exists and needs customer and account context assembled before triage
• The user wants to understand the customer's history and entitlement before responding
• Preparing the context foundation for classification, severity assessment, and response drafting

When NOT to Use

• Creating a new case — use cs-case-intake
• Classifying the issue type — use cs-issue-classifier
• Assessing severity and SLA — use cs-severity-assessment
• Routing the case — use cs-case-routing
• Drafting a response — use cs-response-drafter

Core Instructions

Progress Tracking

Create tasks at the start:

TaskCreate(subject="Locate case and identify customer", activeForm="Locating case details")
TaskCreate(subject="Search for customer account and history", activeForm="Searching for customer data")
TaskCreate(subject="Assemble context packet document", activeForm="Assembling context packet")

Step 1: Locate the Case

Identify the case from the user's message:

• If a Case ID is provided, search for the tracker: SearchM365(sources=["files"], query="service case tracker")
• If a customer name is given, search: SearchM365(sources=["email"], query="<customer name>")
• Read the case details from the tracker using ReadFileContent

Extract the customer name, account ID, channel, and issue summary to guide context searches.

Step 2: Search for Customer Account and History

Run parallel searches across M365:

For each document found, read relevant sections using ReadFileContent.

Step 3: Assemble the Context Packet

Produce a Word document (invoke the docx skill) with:

1. Customer Profile — Name, email, account ID, company, contact details
2. Account and Entitlement Summary — Service tier, contract status, entitlement details, account health indicators
3. Prior Case History — Last 5 cases with dates, issue types, resolutions, and outcomes; flag any repeat patterns
4. Product or Service Context — Products owned, services subscribed, relevant version or configuration details
5. Known Issues and Active Incidents — Any active incidents affecting this customer's product or service area
6. Prior Correspondence Summary — Recent email and Teams threads involving this customer (last 30 days)
7. Account Owner and Escalation Contacts — Internal account owner, escalation manager, executive sponsor (if applicable)
8. Data Gaps — Information not found (e.g., CRM unavailable, no prior cases, entitlement data stale)

Available MCP Tools

Guardrails

• Mask sensitive financial details — credit card numbers, billing specifics, and payment information must not appear in generated documents
• Cite source and retrieval date for every data point from CRM or prior cases
• Flag stale data — if customer entitlement data is older than 30 days, call it out prominently
• Operate in read-only mode — never update CRM, account records, or case status
• Flag data gaps — if CRM is unavailable, prior cases not found, or entitlement data missing, surface it in the Data Gaps section
• Note permission restrictions — if any data sources were inaccessible due to permissions, note this
• Preserve customer privacy — do not include customer data in any artifact beyond what is needed for the case