Skip to main content

Run your entire GTM as one system.

It’s a harness for your whole GTM: it knows your business, works your tools, and does the grunt work so you spend your time on the work only you can do. The more you use it, the more it compounds.

See it work →

you directed:draft outbound to the Meridian Freight buying committee

ACCOUNT BRIEF · Meridian Freight

built fromHubSpot3 callsICP model

Buying committee: 4 named, 2 economic buyers.

Pain (from the Q2 call): margin squeeze on freight lanes.

Lead with our SOC2 badge

you: open on the margin pain, not the badge

Pricing section — waiting on your read

The harness remembers this correction. The next brief starts sharper.

A skill runs the same steps every time. A prompt starts over.

A prompt starts from scratch every run, so what you get back swings. A skill has your method written in, so it runs the same way every time, the way you defined it.

/produce-content is one skill. Here’s what it runs.

1

CONTEXT LOADING (your method, written in)

  • ├── ICP synthesis
  • ├── Voice & tone standards
  • ├── Content principles
  • ├── Content-type mapping
  • └── Recent outputs
2

EXECUTION (calls other skills)

  • ├── calls /persuasive-copywriting
  • ├── calls /generate-image
  • ├── runs the pipeline: Angles → Outline → Draft → Score → Polish
  • └── all of it uses the context from Step 1
3

QUALITY GATES (checks its own work)

  • ├── calls /edit-content
  • ├── calls /skeptical-buyer
  • ├── calls /final-content-reviewer
  • └── voice checked before anything ships

OUTPUT: Blog + LinkedIn + Newsletter + Images

One command, and the same method runs end to end.

Bigger jobs are just skills calling skills. This one calls five.

The Research Agent

It runs on the stack you already have.

One command, and eight agents run in sequence on your CRM: relationship history, company intel, tech stack, pain signals, the buying committee. What comes back is a prospect brief your rep can walk into the call with.

It uses the tools you already pay for, your CRM, the browser, the web, not a walled garden.

HubSpot / SFDC / Attio (MCP)Playwright (MCP)PerplexityApifyFirecrawlWebSearchClaude Vision
Trigger

One command

“research Acme”
Context

Reads your world

It reads your CRM and your past calls before it does anything.
Branches

Splits into specialists

  • ›Company research — news, signals, basics
  • ›Tech-stack discovery — tooling + complements
  • ›Pain-signal mining — tuned to your ICP
  • ›Committee mapping — MEDDPICC roles + influence

…each working in parallel, folding into one synthesis.

Gates

Clears the bar

  • 2+ sources— every finding confirmed
  • >90% match— validated vs. reality
  • 100% cited— URL + timestamp
Artifact

A MEDDPICC-ready brief

The real output, shown — not a mockup.

5,900 customer calls, pulled and read, in one afternoon.

MEDDPICC roles, deal history, contact intel, loaded before you ask.

One agent, one account, start to finish. Or we’ll run it live on our call, on a company you name.

That’s one agent. STEEPWORKS is a system of them.

Spin up data-driven campaigns to test your market.

  1. ICP

    /revops icp build

    from your closed-won deals and call data

  2. Campaign ideation

    /hypothesis

    Find campaign ideas from your data, or tune ones you run

  3. List Build

    /research-prospect

    Discover + rank prospects by fit

  4. Messaging

    /persuasive-copywriting

    Variations across three frameworks

  5. ↺ Re-score

    /hypothesis

    Analyze, iterate, refine

feeds back to ICP — each run starts sharper

Once the market tells you what works, you produce it at scale.

Ship like a content team many times your size, all on strategy.

Four data sources converge into your context, then the skill chain ships it.

Market & competitors
PerplexityApifyFirecrawl
Tracked feeds
65+ RSSCloudflare Worker → Supabase

the voices you follow

Deep research
/aeo-map

gap × relevance

SEO gaps
GSCSerper

serp / semantic-cluster scripts

Context

everything the engine knows about your market, assembled

/produce-content

draft

/edit-content

editorial

/skeptical-buyer

buyer-lens

/persuasive-copywriting

sharpen

/review-content

quality-gate

Blog · LinkedIn · Newsletter · one-pager

one strategy, one voice

Idea → shipped in ~25 min · by hand: 6-8 hrs

one strategy, one voice

Your third correction is your last.

The third time you fix the same thing, it drafts a rule and waits for you to approve it before anything changes.

  1. Edit captured

    your correction, captured as-is

  2. 3× threshold

    the same fix, three times → a pattern

  3. Rule drafted

    the system proposes a rule

  4. You approve

    the review gate — nothing applies until you sign off

  5. Applied

    now it's a permanent rule, loaded every run

captured & proposedyour gate

500 sessions in, it still loads as fast as session one, and it sounds like you.

It never edits itself behind your back. What compounds is a system you approved, rule by rule.

Every workflow that runs in production.

Six pipelines that run in production. Each one an agent orchestrating skills into a finished job.

Research Agent

Full prospect intelligence — one command to a MEDDPICC-ready brief.

See the run →

Content Engine

Idea to publishable draft — angle, draft, edit, quality gates chained.

See the run →

Competitive Intel

A live read on a competitor — research, positioning, battle-card.

See the run →

Newsletter Pipeline

A weekly edition end-to-end — scrape, evaluate, draft, gates, send-ready.

See the run →

ICP Scoring

A target list, ranked — define, enrich, score, deep-dive.

See the run →

Meeting Prep

A call you walk into ready for — CRM, attendee research, dossier.

See the run →

The full library

63 workflows · 49 skills

Workflows (63)

Point it at your first workflow.

Pick one above and see the run, or book a call and we’ll map the one your team needs.

Book a strategy call