B2B SDR Agent

The problem it addresses

B2B export teams often spread sales work across WhatsApp, Telegram, email, and manual spreadsheets. New leads respond slowly, qualification rules stay inconsistent, cross-timezone follow-ups fall through, and quote or closing records are hard to retain. The skill frames this as a continuous lead-to-close workflow rather than single-point auto-reply.

How it works

B2B SDR Agent organizes the sales agent with a 7-layer context system: IDENTITY.md stores company, role, and product catalog; SOUL.md defines tone and communication rules; AGENTS.md describes the 10-stage sales pipeline; USER.md records owner profile, ICP scoring, and admin whitelist; HEARTBEAT.md configures 13 scheduled inspections; MEMORY.md uses a four-layer anti-amnesia protocol; and TOOLS.md connects CRM, channels, and external integrations.

Key capabilities include:
- 10-stage pipeline: cold lead, engaged, qualified, proposal, negotiation, and closed.
- Multi-channel routing: WhatsApp, Telegram, and Email with channel-specific strategies, plus delivery-queue for async delivery and retry.
- Lead and quoting: lead-discovery handles lead search and ICP scoring, while quotation-generator creates PDF proforma invoices.
- Memory and knowledge: chroma-memory stores per-turn conversation memory, and graphify organizes product catalog, customer intelligence, and market research.
- Multi-tenant isolation: uses wireproxy SOCKS5 to a Cloudflare WARP account for an independent egress IP, reducing WhatsApp device-linkage risk across tenants.

Boundaries and caveats

This is a deployment-heavy system. It expects a Linux server, Node.js 18+, an AI model API key, and an OpenClaw gateway. Teams without stable CRM, channel credentials, or ops environments will face higher setup complexity. It fits B2B export teams with a defined ICP, product catalog, and after-sales workflow, but not lightweight one-off campaigns built only around temporary chat tools.