Sales Plugin Capabilities: A Complete Guide to Claude's Sales Automation Tools

The Sales plugin bundles slash-commands like /forecast and /call-summary with intelligent skills for account research and pipeline analysis, operating standalone or super-charged via MCP connectors to eliminate manual sales busywork.

The sales plugin capabilities in the anthropics/knowledge-work-plugins repository transform Claude into a comprehensive sales acceleration platform. Built on the standard knowledge-work-plugin architecture, this modular system combines explicit commands with automatically triggered skills to handle everything from call preparation to competitive intelligence. The plugin functions entirely through declarative markdown files and JSON descriptors that define interactive UI flows, argument schemas, and conditional logic for both standalone and connected modes.

Architecture and File Structure

The plugin follows a strict directory structure that Claude parses at runtime to expose functionality. Understanding this layout reveals how the sales plugin capabilities extend Claude's core behavior.

Plugin Descriptor and Metadata

The entry point resides at sales/.claude-plugin/plugin.json, which defines the plugin name, semantic version (1.2.0), and description. This JSON file is the discovery mechanism that Claude reads to register available commands and skills.

Documentation flows through three critical markdown files:

• sales/README.md – Lists high-level slash-commands and implicit skills, plus installation workflows
• sales/CONNECTORS.md – Catalogs MCP categories (CRM, calendar, chat, enrichment) and placeholder tokens like ~~CRM and ~~calendar
• sales/.claude/settings.local.json – Stores user presets for quota targets and product details that personalize forecasts without repetitive prompting

Skill Declaration Schema

Each capability lives in sales/skills/<skill-name>/SKILL.md following a strict YAML-frontmatter schema:

---
name: forecast
description: Generate weighted sales forecasts with scenario planning
argument-hint: Quarter or time period to analyze
---

This declarative approach allows Claude to generate interactive, multimodal UI elements while parsing conditional branches for standalone versus super-charged execution paths.

Explicit Slash-Commands

The sales plugin capabilities surface three primary slash-commands that users invoke directly through the chat interface.

/call-summary

The /call-summary command processes call notes or transcripts to extract structured action items, draft follow-up emails, and optionally log activities to connected CRMs. According to the source in sales/skills/call-summary/SKILL.md, the skill accepts pasted text or uploaded audio transcripts, then returns a markdown-formatted summary with owner assignments and deadline extraction.

/forecast

The forecasting command generates weighted pipeline projections using best-case, likely-case, and worst-case scenarios. As implemented in sales/skills/forecast/SKILL.md, the skill accepts a quarter identifier (e.g., Q2) and returns:

• Commit versus upside breakdowns
• Gap-to-quota analysis
• Risk flagging based on deal stage aging

The command operates in standalone mode with manual CSV uploads or super-charged mode by pulling live opportunity data through ~~CRM connectors.

/pipeline-review

Running /pipeline-review triggers a comprehensive health analysis that scores each opportunity, identifies stale deals surpassing threshold days, and outputs a prioritized weekly action plan. The underlying skill in sales/skills/pipeline-review/SKILL.md supports both uploaded pipeline exports and direct CRM queries via MCP integrations.

Automatic Skills (Intent-Triggered)

Beyond explicit commands, the sales plugin capabilities include domain-specific skills that Claude invokes automatically when detecting sales-related intents in conversation.

account-research

Triggered by queries about prospects or companies, this skill executes web searches and enriches findings via data-enrichment MCPs like Clay or ZoomInfo. It returns consolidated briefings on company metrics, recent news, and key decision-maker backgrounds.

call-prep

When users mention upcoming meetings (e.g., "Prep me for my call with Acme Corp tomorrow"), Claude invokes the call-prep skill defined in sales/skills/call-prep/SKILL.md. The skill gathers account context, retrieves attendee biographies from LinkedIn or CRM data, and suggests agenda items based on historical email threads.

daily-briefing

The daily-briefing skill aggregates calendar events, pipeline alerts, and email priorities into a morning dashboard. When super-charged with ~~calendar and ~~email connectors, it pulls live meeting transcripts and flags urgent deal communications requiring immediate response.

draft-outreach

This skill researches a prospect through account-research sub-tasks, then generates personalized email and LinkedIn message variants. The workflow targets specific personas (e.g., "VP of Engineering at TechStart") and adapts tone based on industry verticals discovered during enrichment.

competitive-intelligence

Invoked when users mention competitors, this capability builds comparison matrices pricing intelligence, and differentiation talk-tracks. It scans the web for recent product announcements and pricing changes, then structures the output into battle cards suitable for sales enablement decks.

create-an-asset

The create-an-asset skill in sales/skills/create-an-asset/SKILL.md generates custom sales collateral including landing pages, presentation decks, one-pagers, and workflow demos. It prompts for audience specifications (e.g., "Finance team at GlobalBank") and value propositions, then returns polished outlines ready for PDF generation.

MCP Connector Architecture

The sales plugin capabilities achieve tool-agnostic flexibility through placeholder tokens defined in sales/CONNECTORS.md. Rather than hardcoding Salesforce or HubSpot APIs, skills reference abstract categories:

• ~~CRM – Opportunity data, account records, activity logging
• ~~calendar – Meeting schedules, transcript retrieval
• ~~email – Thread analysis, communication history
• ~~enrichment – Third-party data providers (ZoomInfo, Apollo, Clay)

A runtime .mcp.json configuration maps these placeholders to concrete MCP servers, allowing the same skill code to function across different technology stacks without modification.

Standalone vs. Super-Charged Modes

Every skill guarantees standalone functionality using only user-provided inputs. For example, the forecast skill accepts pasted CSV deal data and manual quota entry when no CRM connects.

Super-charged mode activates when MCP integrations supply live data, augmenting responses with:

• Real-time opportunity probabilities from CRM stage history
• Calendar-confirmed meeting attendance versus scheduled invites
• Enrichment-driven personalization tokens in outreach drafts

This dual-mode architecture ensures zero-setup usability while rewarding integration with richer context.

Configuration and Personalization

The sales/.claude/settings.local.json file stores persistent metadata that streamlines repeated workflows:

{
  "quota": 1000000,
  "product_details": "AI-powered sales automation",
  "territory": "North America"
}

When present, the /forecast command automatically references these values for gap analysis, and create-an-asset pulls product descriptions without requiring re-entry.

Summary

• The Sales plugin in anthropics/knowledge-work-plugins provides seven automatic skills and three explicit slash-commands (/call-summary, /forecast, /pipeline-review) for end-to-end sales workflows.
• Declarative SKILL.md files in sales/skills/<name>/ define UI flows, argument hints, and conditional logic for standalone or MCP-super-charged execution.
• Tool-agnostic architecture uses placeholder tokens (~~CRM, ~~calendar) mapped via MCP configuration, enabling integration with any connector implementing the category interface.
• Dual-mode operation ensures every capability works without integrations while enhancing outputs when connected to live data sources.
• Personalized settings in .claude/settings.local.json eliminate repetitive data entry for forecasts and asset generation.

Frequently Asked Questions

What is the difference between slash-commands and skills in the Sales plugin?

Slash-commands are explicit triggers (like /forecast) that users type to invoke specific workflows, while skills are automatically detected capabilities that Claude invokes based on conversation context (such as preparing for a call when you mention an upcoming meeting). Both use the same underlying SKILL.md definitions in the file structure, but slash-commands offer direct access whereas skills operate implicitly to assist throughout natural conversations.

Do I need a CRM integration to use the Sales plugin capabilities?

No. Every skill operates in standalone mode using manual inputs like pasted CSV files or typed descriptions. The ~~CRM and other MCP placeholders are optional enhancements; when connected, they super-charge skills with live data, but the core functionality—including forecasting, call summaries, and content creation—works entirely offline with user-provided context.

How does the forecasting skill calculate weighted projections?

The /forecast skill calculates weighted forecasts using probability-adjusted opportunity values based on sales stage percentages, then generates best-case, likely-case, and worst-case scenarios. According to sales/skills/forecast/SKILL.md, the algorithm applies standard weighting to pipeline stages (e.g., 90% for closed-won, 50% for negotiation) and compares the aggregate against quota targets stored in settings.local.json to produce gap analysis and risk flags.

Can I customize the collateral templates generated by the create-an-asset skill?

Yes. The create-an-asset skill accepts explicit parameters for audience, value propositions, and branding preferences through interactive prompts defined in sales/skills/create-an-asset/SKILL.md. While the base templates follow standard sales formats (one-pagers, decks, landing pages), the skill adapts content structure based on your inputs and any product details stored in the local settings file, producing customized outlines rather than rigid boilerplate.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}