An open-source operating system for B2B account-based marketing and enterprise go-to-market, built as Claude Code skills.
Run a real ABM program from the command line: size the market, define and score the ICP, source and tier the account list, map the buying committee, wire the signal engine, build the plays, write the cadences and the air cover, align sales, and measure the whole thing against pipeline instead of impressions.
No app to host. No account. No database. Twenty-three skills plus plain text in a workspace/ folder on your machine.
Most ABM content is either vendor marketing for a platform you have not bought, or a conference talk about "1:1, 1:few, 1:many" that stops before anyone has to build anything. The operating detail is where programs die: how the tier thresholds get calibrated, who is allowed to be contacted at an account and who is a gatekeeper, which signal fires which play, what marketing owns versus what sales owns after the handoff, and what number you put on the board at the QBR.
ABM & Scale OS is that operating detail, written down as executable skills. It comes out of running enterprise ABM programs inside B2B companies, not out of a platform's best-practice PDF.
Three beliefs are baked into every skill:
- Account selection is the program. A badly chosen list cannot be rescued by good creative. More of the work belongs at the front of the funnel than most teams put there.
- Multi-threading is the mechanic that moves enterprise deals. Single-threaded accounts stall. Every play in here is designed to add stakeholders, not just touches.
- Air cover without outbound is advertising. Ads, events, and content earn their keep when they are synchronized to a named account list that a human is also working.
Foundation
abm-program-setup– program charter, objective, budget, motion, and the workspace it all writes toicp-definition– three-dimensional ICP (firmographic, trigger, behavioral) with a 10-point account scoring rubricaccount-sourcing– size the TAM, source the account universe, dedupe, suppress, and hand back a clean listaccount-tiering– Fit plus Intent into a composite score, calibrated to your list's real distribution, then T1/T2/T3
Account intelligence
buying-committee-map– the roles that decide, the ones that block, and the multi-threading rules per tieraccount-research– a Tier 1 account dossier built from public evidence, with a research prompt you can run at scalesignal-engine– the trigger taxonomy, recency decay, scoring, and the signal-to-play routing tablewhitespace-expansion– the installed-base motion: new business units, new stakeholders, product whitespace
Narrative and assets
positioning-narrative– Dunford's five steps run against enterprise buyers, including the do-nothing alternativemessaging-hierarchy– the order you say things in, validated by the frontline, per persona and per segmentbattlecards– competitive alternatives, win themes, objection handling, and the traps to set
Plays
play-designer– 1:1, 1:few, and 1:many play architecture with entry criteria, exit criteria, and a kill ruleoutbound-cadence– multi-channel cadences by segment, with a call and a social touch behind every emailair-cover-ads– account-matched paid social, thought leader ads, audience sync, and suppressionfield-and-events– executive dinners, trade shows, and roadshows, with the pre, at, and post motion that makes them payexec-engagement– executive sponsor programs, one-to-one briefings, and executive business reviews
Alignment and operations
sales-alignment– the SLA, the joint account plan, the handoff definition, and rep enablementdata-orchestration– the stack, the routing logic, suppression, hygiene, and the adapter layer for tool swapsenterprise-deal-support– mutual action plans, security review, procurement, and the late-stage marketing motion
Measurement
abm-measurement– account engagement scoring, coverage, influence, the QBR, and the metrics that are lies
Quality
writing-quality– the anti-slop gate every writing skill runs before it returns copyartifact-eval– scores an artifact against a rubric and revises until it passes
Operate
abm-and-scale-os– the front door: program status, next best action, and the weekly operating cadence
Full stage-by-stage map: docs/lifecycle.md.
- Install the skills (see INSTALL.md). Copy the
skills/folders into your Claude Code skills directory, or clone and point Claude Code at them. - Set up the program. Run
abm-program-setup. It interviews you on objective, motion, budget, and constraints, then writesworkspace/program/. - Build the list. Run
icp-definition, thenaccount-sourcing, thenaccount-tiering. You end with a scored, tiered, suppressed target account list inworkspace/accounts/target-accounts.csv. - Start the loop. Run
abm-and-scale-os. It reads the program state and tells you the next move, whether that is research on a Tier 1 account, a play that needs building, or a cadence that has gone cold.
- No invented numbers. Nothing fabricates a stat, a customer, a case study result, or an intent score. If a number is not in your source material, the skill asks for it or marks it unknown.
- Human in the loop. Nothing sends, publishes, or launches. Skills draft, score, and recommend. You approve.
- Tier discipline. Tier 1 effort is expensive and finite. Every skill enforces the difference between what a T1 account gets and what a T3 account gets.
- Sales is a co-author, not a recipient. Account plans, messaging, and handoff rules are built with the reps who will run them, and every skill says where that review happens.
- Local and private. Your account lists, dossiers, and program docs live in
workspace/on your machine..gitignorekeeps them out of git by default.
The skills are written against a specific stack so they are runnable on day one, with substitutions documented for every layer:
| Layer | Reference | Common swaps |
|---|---|---|
| List building and enrichment | Clay | Apollo, ZoomInfo, Ocean.io, Sales Navigator |
| CRM | HubSpot or Salesforce | Pipedrive, Attio |
| Sequencer | Salesloft or Outreach | Apollo, Smartlead, Instantly, HubSpot Sequences |
| Marketing automation | HubSpot, Marketo, or Pardot | Customer.io, Braze |
| Paid air cover | LinkedIn Campaign Manager | Meta, StackAdapt, programmatic display |
| Intent and signals | Predictleads, UserGems, Champify | 6sense, Demandbase, Bombora, G2 Buyer Intent |
| Research | Claygent or a research agent | Perplexity, manual analyst work |
Full detail, including the "you have no budget" path for every layer: docs/stack.md.
Issues and PRs welcome. Good first contributions: a new play in docs/plays-library.md, a vertical-specific ICP rubric, an adapter for a tool not in the stack table, or a new artifact-eval spec. House rules: no invented numbers, human in the loop, and every recommendation has to survive the question "does this move pipeline?"
MIT. Fork it, adapt it, run it inside your own company. If it books you a meeting at an account you could not get into before, that is the point.