Sales Pipeline
The own-stack RevOps operator. Teaches an OPVS/OpenClaw agent to run the CRM the stack already owns — the AgentBoard deals/contacts/companies boards — through the full deal lifecycle (new lead → qualify → demo → proposal → negotiation → won/lost): compose a lead dossier from SpiderPeople/SpiderSite/IDAP + icp-discovery, trigger SpiderFlow automation on stage changes, progress stages only on logged buyer signal, run a Cardone-grade fast+persistent follow-up cadence, close + feed win-loss to agentmemory, and emit stage-velocity KPIs. Wraps NO third-party CRM (AgentBoard IS the CRM); ships NO backend. Backed by 81 evidence-bundled registry assets (incl. Belfort Straight-Line + Cardone closing/follow-up methods, re-expressed) + 23 learnings + 5 enforcement scripts. guidance-only meta-package: the agent's own LLM operates the CRM following SKILL.md; state + IO live in the declared peers.
Most AI CRM tools advance your deals for you. Mine refuses to.
I built sales-pipeline after my own agent moved a deal to Proposal off the words "let me think about it." Nobody had agreed to anything. A pipeline that only moves forward cannot forecast, so the rule here is one sentence: a deal changes stage only when there is a logged buyer signal, dated at or after the deal entered its current stage. No signal, no move. A script checks every transition and prints PASS or FAIL with a count.
It wraps no third-party CRM. The AgentBoard you already run is the CRM: columns are stages, a card is a deal, custom_fields is the deal data. There is no sync and no second source of truth.
Worked example: a reply lands, the agent creates the deal, composes a dossier from your enrichment peers, scores it, and stops at Qualified until the dossier passes coverage. It only reaches Proposal when a real buyer signal exists.
Guidance-only. 12 references, 9 recipes, 81 registry assets, 23 learnings, 5 scripts, zero methods. Every write is done by a peer you already have.
Features5
Advances a deal only on a logged buyer signal, and a script proves it
Uses your AgentBoard deals, contacts and companies. No third-party CRM
Composes ICP fit, company and persona context from your enrichment peers
Drives a fast, persistent cadence off the deal's own touch dates
Keeps the process, drops urgency and absolute claims for REACH and CLP
Videos1 video
DemoSales Pipeline · Demo video
Demo · SpiderMediaArticles about Sales Pipeline2 articles
I thought my sales AI was closing deals. It was guessing.
opvs.ai22 Jul 2026I thought my sales AI was closing deals. It was guessing.
A prospect said "let me think about it" and my agent moved the deal to Proposal. Nobody had agreed to anything. Here's the rule that fixed it: a deal moves stage only on a logged buyer signal, enforced by a script that fails the board when it can't find one.
Videos1 video
DemoSales Pipeline · Demo video
Demo · SpiderMediaArticles about Sales Pipeline2 articles
Showing all 2
SKILL.mdthe file your agent reads
SKILL.md loads when you open this tab.
What's new4 releases · newest first · from the package's CHANGELOG
v0.1.42026-11-04 — marketplace launch filmLatest
### Added - Marketplace launch video (`media.demo_video_url`) on the skill listing.
v0.1.32026-07-22 — customer-visible + clean bundle
### Changed - **Tier is now `beta`.** 0.1.0 through 0.1.2 published at `internal`, which hides a package from customers. The skill is now visible on the marketplace and installable. - **`skills[0].visibility` is now `public`** so the skill entry resolves for installers, not just the package. - **The published bundle contains only the skill.** Staged from the five canonical entries plus `skills/sales-pipeline/`. Earlier bundles shipped `BRIEF.md`, `tests/`, `creatives/`, `marketing/` and `propagation/` to customers. ### Added - Marketplace card assets: `media.hero_image_url` + 3 screenshots (durable `media.spideriq.ai` URLs), `marketing.long_description`, 5 `marketing.features`, and a `marketing.author` block. Previous versions rendered a blank listing. - `type: "skill"` on the manifest. ### Notes - **No behaviour change.** The router, 12 references, 9 recipes, 81 registry assets, 23 learnings and 5 enforcement scripts are unchanged from 0.1.2.
v0.1.12026-06-18 — substrate alignment (re-pointed at the real platform primitives
### Changed — composes the real capabilities instead of thin/assumed ones - **Agent owns the board** — `agentboard-crm-anatomy` rewritten: the agent provisions the deals board, creates/renames columns to its pipeline, sets per-column + per-deal instructions + subtask handoffs, JSONB custom_fields, and configures `heartbeat` (cadence scheduler) + `soul` (deal persona). Kills the v0.1.0 "fixed stages" assumption. - **Cadence engine = heartbeat + SpiderMail + smartlead** — `followup-cadence-engine` rewritten. v0.1.0 assumed a non-existent SpiderFlow cadence flow; the real scheduler is the OpenClaw/board heartbeat. No phantom backend. - **Lead-gen + qualify = the campaign flow + CHAMP** — `spiderflow-trigger-catalog` rewritten around `campaign_maps_site_verify` (single/batch/campaign: Maps→site→leads→**CHAMP score**→verify→CRM/IDAP→smartlead-export). `qualify-and-dossier` now consumes the flow's `champ_analysis` instead of hand-rolling enrichment. `new-lead-arrival` gains the "get me leads for this ICP" campaign variant. - **Stage model** — script `verify-stage-progression-evidence.mjs` now honors the board's actual columns (`stage_order` override); SKILL.md contract aligned to real OPVS `deals` defaults (Discovery/Proposal/Negotiation/Contract/Won/Lost) + the skill's fuller pipeline. Fixed the v0.1.0 hardcoded-stage defect. ### Added - **`references/spiderpublish-personalized-lp`** — personalized landing page per deal from IDAP (SpiderPublish), at demo/proposal stage. Recipes `demo-booked-prep` + `proposal-stage` updated. - Peer `@spideriq/publish-skills` (SpiderPublish). ### Why Verified against the real OPVS/SpiderIQ source: the platform already does CHAMP scoring, campaign lead-gen, heartbeat scheduling, and personalized LPs. v0.1.0 under-composed it. v0.1.1 points the skill at the real primitives.
v0.1.02026-06-17
### Added — first release (greenfield, guidance-only, v2.1 / v2.2) - **SKILL.md router** — signal-in → stage-routing decision tree; 6 hard rules; v2.2 coordination block (all 4 `/#` cross-refs); the deals-board pipeline contract. - **11 references** — agentboard-crm-anatomy · spiderflow-trigger-catalog · lead-dossier-composition · pipeline-stage-frameworks · stage-transition-decision-logic · followup-cadence-engine · win-loss-memory-pattern · compliance-pipeline-vocabulary · multilingual-stage-localization (v0.2 stub) · handoff-to-retention · handoff-to-content. - **9 recipes** — new-lead-arrival · qualify-and-dossier · demo-booked-prep · proposal-stage · negotiation-stage · closed-won-handoff · closed-lost-debrief · compliance-pipeline · multilingual-pipeline (v0.2 stub). - **81 registry assets** — stage frameworks (MEDDIC/SPICED/BANT), 2025 conversion benchmarks, win-loss question banks, REACH/compliance rule cards, stage-anatomy blueprints, scrapecreators-sourced teardowns, and the re-expressed Belfort Straight-Line (Three-10's exit gate, 4-archetype classifier, looping, A-Lot funnel) + Cardone (5-min follow-up rule, 5th-12th curve, always-proposal, 3-options, second sale, objection taxonomy) method assets. - **23 learnings** — incl. external-crm-temptation anti-pattern, phantom-stage-progression, follow-up-is-highest-leverage, cardone-prescribes-this-skill, and the Belfort/Cardone IP+ethics guardrails. - **5 enforcement scripts** — verify-dossier-coverage · verify-stage-progression-evidence · lint-external-crm-mentions · verify-handoff-payload · verify-coordination (PASS/FAIL with numbers). - **instructions/** — AGENTS.md (non-Claude runtime mirror) + DEPLOYMENT.md (how deal output feeds retention + content-team). ### Notes - guidance-only: NO schema.yaml, NO backend; every IO is a declared peer's. - EN floor; DE/RU/FR/RO/HE/ES planned for v0.2.0. - SpiderFlow peer package name (`@spideriq/core-skills`) confirmed at publish-time peer audit.
Showing all 4
