CRM Customer Management Skill

Enterprise CRM workflow orchestration for AI agents — customer 360 lookups, deal pipeline management, activity logging, and multi-MCP onboarding across Salesforce, HubSpot, Zoho CRM, and Pipedrive.

What This Skill Does

This skill turns raw CRM tool access into reliable, repeatable workflows. Instead of figuring out which of the 28 CRM tools to call and in what order, the agent follows proven recipes:

Without this skill:

• Agent makes 10+ trial-and-error tool calls
• Inconsistent output format every time
• Forgets to check for duplicates before creating
• No audit trail (notes/activities not logged)
• Doesn't know how to combine CRM with email/calendar

With this skill:

• Deterministic tool sequences (3-5 calls per workflow)
• Consistent report templates every time
• Duplicate detection built into every create workflow
• Automatic audit trail on every write operation
• Cross-MCP orchestration (CRM + Email + Calendar + Slack)

Installation

Claude.ai

1. Download this folder as a ZIP
2. Go to Settings > Capabilities > Skills
3. Upload the ZIP

Claude Code

# Clone into your skills directory
git clone https://github.com/zavora-ai/skill-crm-customer-management.git \
  ~/.skills/skills/crm-customer-management

ADK-Rust (adk-skill)

Place in your project's .skills/skills/ directory:

cp -r crm-customer-management /path/to/project/.skills/skills/

The adk-skill crate will auto-discover and score it against user queries.

Requirements

Required:

• mcp-crm server connected and authenticated (Salesforce, HubSpot, Zoho, or Pipedrive)

Optional (for cross-MCP workflows):

• mcp-email — send follow-up emails after activities
• mcp-calendar — schedule meetings when deals advance
• mcp-slack — notify team on deal wins/risks
• mcp-notifications — alert owners on stalled deals

Folder Structure

crm-customer-management/
├── SKILL.md                    # Main skill file (loaded when triggered)
├── scripts/
│   ├── validate_contact.py     # Data quality validation before creates
│   └── score_pipeline.py       # Pipeline health scoring + staleness detection
├── assets/
│   ├── customer-360-report.md  # Output template for customer lookups
│   └── pipeline-review-report.md # Output template for pipeline reviews
├── references/
│   ├── tool-sequences.md       # Exact tool call patterns (loaded as needed)
│   ├── cross-mcp-workflows.md  # Multi-server orchestration recipes
│   └── examples.md             # 5 real-world scenarios with traces
├── README.md                   # This file (for humans, not loaded by agent)
└── LICENSE

How It Works

Progressive Disclosure

1. Always loaded (~50 tokens): Name + description in frontmatter — enough for the agent to know when to activate
2. Loaded on trigger (~800 tokens): SKILL.md body — decision tree, workflow steps, constraints
3. Loaded as needed: References (tool sequences, examples) — only when the agent needs deeper guidance

Decision Tree

User request arrives
├── Person name/email? → Customer 360 Lookup
├── "deal", "pipeline", "forecast"? → Deal Management
├── "call", "meeting", "log"? → Activity Logging
├── "onboard", "new customer"? → Customer Onboarding
├── "pipeline review", "health"? → Pipeline Review
└── Unclear? → Ask for clarification

Example Usage

User: "What's the status of our deal with Acme?"

Agent behavior:

1. Routes to Deal Management workflow
2. Calls search_deals(query: "Acme") → finds deal
3. Calls get_deal(id) → gets stage, value, close date
4. Calls list_deal_contacts(deal_id) → shows who's involved
5. Calls list_activities(record_id, limit: 5) → recent engagement
6. Formats response using consistent structure

Result:

## Acme Corp Enterprise Deal
Stage: Negotiation | Value: $75,000 | Close: Feb 28
Contacts: John Smith (Decision Maker), Lisa Chen (Technical)
Last activity: 3 days ago ✅

See references/examples.md for 5 complete scenarios.

Success Criteria

Scripts

validate_contact.py

Deterministic validation before creating contacts. Checks email format, phone format, name length, and flags personal email domains.

python scripts/validate_contact.py '{"name": "John Smith", "email": "[email protected]", "phone": "+1-555-0123"}'
# → {"valid": true, "errors": [], "warnings": []}

score_pipeline.py

Calculates weighted pipeline value, identifies stalled/at-risk deals, and produces a health score (0-100).

python scripts/score_pipeline.py '{"deals": [...], "quota": 500000}'
# → {"weighted_value": 234000, "health_score": 72, "at_risk": [...]}

Contributing

1. Fork this repo
2. Make changes to SKILL.md or references
3. Test with 10+ queries that should trigger the skill
4. Test with 5+ queries that should NOT trigger it
5. Submit PR with test results

Related Skills

• itsm-service-management — IT service desk workflows
• customer-service-operations — Support ticket management
• payment-processing — Governed payment workflows

MCP Server Compatibility

This skill is designed for the mcp-crm server (v1.1.0+):

All 28 tools across contacts, companies, deals, activities, pipelines, notes, and associations are orchestrated by this skill.

Contributors

License

Apache-2.0

Part of the ADK-Rust Enterprise skills ecosystem.

Built with ❤️ by Zavora AI