SandsDX

GTM agents and skills

These are the agents, skills and templates I use to build pipeline for B2B SaaS companies. Each one answers a question a CEO, CRO or operating partner asks in the first 30 days, and each one runs in Claude Code.

They are listed in the order a revenue team would use them. 5 of 10 are in use today. The rest are in build and will be published as open source with a sample output on synthetic data.

Set the context

Agents are only as good as what they know about the company and the rules they run under.

Bravenger

Live

Do our agents know what the company believes and how it decides?

A company brain template: brand, ICP, personas, positioning and messaging as versioned Markdown files that every agent reads before it writes. Lint scripts and a test suite check the brain in CI on every change.

Produces
A forkable company brain with a worked example and governance rules for agents.
Form
Template
Access
Open source (MIT)
GTM layers
  • Foundation

GTM Harness

In build

How does a revenue team run Claude Code every day without losing control of it?

The operating setup around the agents: skills for recurring GTM work, hooks that block unsafe actions, connectors to the CRM and call recorder, and written rules the agents follow. It bundles the other tools on this page.

Produces
An installable harness for a revenue team, with a starter set of skills and guardrails.
Form
Claude Code harness
Access
Open source
GTM layers
  • Foundation
  • Motion

Find the leaks

Diagnose where pipeline is lost before changing anything.

New Logo Growth

Live

Where are we losing new-logo pipeline?

Audits a B2B SaaS company from public data across six layers in order: ICP, audiences, messaging, offers, motion and measurement. Scores each layer, ranks the priorities by pipeline effect and backs each play with third-party research checked by hand.

Produces
A one-page pipeline teardown and a full report with 3 to 5 ranked plays.
Form
Claude Code plugin
Access
Open source
GTM layers
  • ICP
  • Audiences
  • Messaging
  • Offers
  • Motion
  • Measurement

Speed-to-Lead Test

Live

When a buyer asks for a demo, how fast and how well do we respond?

Runs one demo request through your own funnel, the way a buyer would, and logs every email, call, text and LinkedIn touch for seven days. Each reply is scored for time to first human touch, number of touches, personalization and whether it books a meeting. I also use it on a single prospect before reaching out.

Produces
A scorecard showing what your buyer experienced, minute by minute, with the fixes ranked by points gained and a model of the pipeline a faster response would add, built on your own funnel numbers.
Form
Claude Code plugin
Access
Open source (MIT)
GTM layers
  • Motion
  • Measurement

Win/Loss Analyzer

In build

Why do we actually win and lose?

Reads call transcripts and CRM closed reasons together, separates the stated reason from the reason the calls show, and groups the patterns by competitor, segment and stage.

Produces
Win and loss themes with quoted evidence, competitor patterns and the message changes they point to.
Form
Claude Code skill
Access
Open source
GTM layers
  • Messaging
  • ICP

Aim and engage

Pick the right accounts, reach them at the right moment, and tell one story when you do.

ICP Fit Scorer

In build

Which accounts should we chase, and which good-fit accounts are we missing?

Learns the pattern of closed-won accounts from firmographic and technographic data, scores the current account list against it and finds high-fit companies that are not in the CRM.

Produces
Tiered account list, whitespace list and the scoring method as a config file.
Form
Claude Code skill
Access
Open source
GTM layers
  • ICP
  • Audiences

Signal Outbound Agent

In build

Which accounts are in market this week, and what do we say to them?

Watches for buying signals such as hiring, funding and technology changes, scores fit and intent, drafts the message from the trigger and checks every claim against the evidence. A person approves before anything is sent.

Produces
A weekly list of scored accounts with drafted, evidence-checked messages.
Form
Recurring agent
Access
Open source
GTM layers
  • Motion
  • ICP

msgdrift

Live

Do our pages still tell the story we repositioned to?

Compares a website’s live copy with the positioning document that should govern it, and reports where the copy contradicts, dilutes or leaves out the message. It reads for meaning, so matching words alone do not count as on-message.

Produces
An alignment score from 0 to 100, findings per page with severity, and a shareable report.
Form
Web app
Access
Free, no login
GTM layers
  • Messaging

Measure and forecast

Know how much pipeline the plan needs, whether the numbers can be trusted, and where AI puts you on the buyer’s shortlist.

Pipeline Forecast Model

In build

How much pipeline do we need, from which sources, and by when?

Works back from the bookings target through stage conversion, deal size and cycle length to the pipeline each source must create. It checks CRM hygiene, stage definitions and source tagging first, then runs a Monte Carlo forecast of the quarter.

Produces
A pipeline requirement by source and month, a data integrity report and a range of likely quarter outcomes.
Form
Claude Code skill and model
Access
Open source
GTM layers
  • Measurement

Shortlist Monitor

In engagements

When buyers ask AI for a shortlist, are we on it, and where?

Asks ChatGPT, Claude, Gemini, Perplexity and other models the same buyer questions on a schedule, several runs per question, and tracks mentions, rank and the sources cited. Gaps become drafted content to close them.

Produces
Shortlist share and rank by model and question, cited sources and gap content drafts.
Form
Recurring agent
Access
Run for clients
GTM layers
  • Measurement
  • Messaging

Building pipeline with these

The tools find the problem and do the repeatable work. Deciding what to fix first, rebuilding the system and getting the team to run it is the engagement. If you are hiring a growth leader or need pipeline built, let's talk.

Schedule a Conversation