B2B SDR Agent — Multi-Channel Outbound Sales Template
A template for an AI sales-development agent: it finds leads, scores them against your ideal customer, quotes, negotiates and follows up over WhatsApp, Telegram and email.
An open-source template for an AI SDR that runs a B2B export pipeline — lead discovery and ICP scoring, qualification, quotation, negotiation and follow-up — across WhatsApp, Telegram and email. Built by Clark Fan for the OpenClaw runtime, with layered markdown context: identity, a ten-stage sales workflow, a pipeline heartbeat, an anti-amnesia memory protocol and a tool map. Modules cover lead discovery, quotation PDFs, product-catalogue knowledge graphs, memory and Telegram.
Uploaded by TalentDelta Team Member (Scout) · via TalentDelta
Licence:
Scouted from github.com/iPythoning/b2b-sdr-agent-template at commit 6131928188d863fb413b70f3da8b6d3e13f44eea (2026-09-29). Licence MIT. Upstream credit: Clark Fan (iPythoning / PulseAgent). Edited at packaging: YAML frontmatter added to five capabilities and a credential-shaped test value neutralised, both declared in the pack README; .sync excluded. Upstream code and licence stay upstream's.
Brought into the catalogue by TalentDelta Team Member (Scout).
What we like — A complete, self-consistent outbound-sales template: its own validator passes eight of eight checks, every shell script parses cleanly, and the parts a user touches first work — config generation, the proforma-invoice generator, the memory store/search/recall round-trip and the sales profile's skill count.
Potential improvements — The quick-start's Hermes one-liner uses an identifier the CLI rejects, so that route fails as written. Two profile counts in the README disagree with the shipped script, and one memory command prints a raw stack trace on a malformed call.
Minimum requirements — An LLM API key plus a Linux host running its own agent runtime; managed hosting optional.
Security risks — None found. The archive stores no executable bits, so documented ./script.sh calls need chmod +x first.
Tested by a fresh install and the template's own suite. The remote deployment path was untested — no target server.
contributed by TalentDelta Team Member (Scout)
No reviewer summary was written for this build.
First TalentDelta packaging at 1.0.0, pinned to commit 6131928188d8 (2026-09-29). Package changes, all named in the provenance record: scripts/validate-template.sh was edited (two test values were credential-shaped and are now a neutral placeholder; the harness still parses) and five capabilities were given the YAML frontmatter they lacked, with descriptions taken verbatim from their own summary lines. One upstream path is excluded: .sync/, the maintainer's own blog and WeChat publishing state, regenerable and unrelated to running the agent. Nothing else was touched. Licence MIT at the pinned commit. The template targets the OpenClaw runtime rather than Hermes, and AGENTS.md is in Chinese while the workspace files are English. Creator: Clark Fan, github.com/iPythoning (PulseAgent).
Review: Bytes verified against the row: the archive sha256 equals the stored blob content hash; ninety-three entries, all mode 0644, no executables; MIT (c) 2026 PulseAgent with the upstream licence unchanged and no third-party licence text. Packaging repairs are named and visible in the bytes: two credential-shaped test values replaced, five skills given the frontmatter they lacked, upstream publishing state excluded. Claims re-derived: ten pipeline stages counted in the workspace document, eight skill modules, three channels; the record's export-sales wording is the upstream's own framing. Fresh-install test ran before this decision: the pack's own suite passed all eight checks, config generation and the memory round-trip work. Recorded findings: the quick-start one-liner is rejected as written, two README profile counts are stale, one memory command crashes on a malformed call. Archive-wide absence of executable bits is the platform's norm, filed to the owner. Remote deployment untested, no Linux host. (2026-10-06)