bama
BAMA: 모든 Mac 앱을 만들 수 있습니다
GT마케팅을 위한 오픈소스 스킬
Open source skills for GTM
스킬TypeScript · JavaScriptGitHub @eliasstravik직접 설치운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
Run your full GTM motion from one shared context
GTM Skills lets a GTM teammate keep one organization's facts, ideal customer profiles, and personas in a git-backed Markdown workspace, qualify prospects against them in conversation, and turn them into saved workflows that run locally or on Vercel.
✓ 100% free and open source ✓ One git-backed workspace ✓ Runs locally or on Vercel
Keep organization, market, buyer, and workflow knowledge in one placeREADME 전문 보기 ↗The GTM workspace records what the organization knows about itself and its team. ICPs define the companies it sells to, personas define the people it sells to, and saved workflows turn that context into repeatable work. Every change is proposed in plain language, approved through your agent's normal write permission, and committed to git, so the history of what the team decided is always there.
| GTM Skills | Repeated prompts | Standalone templates | Custom agents | |
|---|---|---|---|---|
| Installs into Claude Code, Codex, Cursor, OpenCode, and other skills.sh hosts with one command | ✅ | ✅ | ❌ | ❌ |
| Keeps organization facts, ICPs, and personas as Markdown in one git repository | ✅ | ❌ | ❌ | ❌ |
| Proposes every durable change in plain language before writing it | ✅ | ❌ | ❌ | ❌ |
| Checks a pasted list of people or companies against your ICPs and personas without writing anything | ✅ | ❌ | ❌ | ❌ |
| Gives every workflow a typed result table with a stable key, cost, and error per row | ✅ | ❌ | ❌ | ❌ |
| States the cost before every run and offers a one-row test before a new workflow runs at scale | ✅ | ❌ | ❌ | ❌ |
| Caps every run by rows and spend | ✅ | ❌ | ❌ | ❌ |
| Runs the same workflow file on Postgres locally and hosted on Vercel with Neon | ✅ | ❌ | ❌ | ❌ |
| Reaches people in Slack for approvals while a hosted run is in progress | ✅ | ❌ | ❌ | ❌ |
Keep durable GTM knowledge and reusable automations in one repository, and let the agent you already use do the writing.
Say "create an ICP for lean B2B SaaS" or "create a persona for revenue leaders". The agent drafts the definition from your facts and public sources, shows it, and saves it to the workspace only when you approve.
Paste a list of people or companies and ask "are these a fit". The agent checks each one against the saved ICPs and personas and returns a verdict per row. Nothing is written.
Say "build a workflow that scores our inbound companies". The agent writes the workflow and its result table, opens a diagram of what it does and what it costs, and runs it on your say-so, locally or on Vercel, on a schedule when you want one.
Each skill owns one part of the workspace and hands off to the others. Install gtm-workspace in every case; the rest depend on it. Not sure where to begin? Say "how do I get started" or run /gtm-get-started.
| Skill | Owns |
|---|---|
gtm-get-started | Where to begin: looks at the workspace and walks through each missing step with the skill that owns it |
gtm-workspace | The workspace itself: the organization record, members, and workspace health |
gtm-icp | Ideal customer profiles: the companies the organization sells to |
gtm-persona | Personas: the people the organization sells to |
gtm-qualify-prospects | In-conversation fit checks of supplied people or companies |
gtm-workflow | Saved workflows: code, tables, runs, schedules, and deploys, and workflow recipes such as network enrichment |
gtm-agent | GTM Agent itself: deploying the Slack agent to Vercel, its workflow project, and their wiring, from one prompt |
No. The agent writes the workflow, its table, and its migration. You read the diagram, approve the cost, and look at the rows; the code is there when you want it.
Any skills.sh host: Claude Code, Codex, Cursor, OpenCode, and others. One install command puts the skills where every host finds them.
In ~/.gtm/<org-slug>/ on your computer, as Markdown with git as its memory. When you want to share it, the agent creates a private GitHub repository named gtm-<slug> and pushes to it.
Locally: Node; the scaffold brings its own Postgres and starts it with the workflow runtime. Hosted: a Vercel project connected to the workspace repository, a Neon database added through Vercel, and an AI Gateway key. Each workflow declares its row and spend caps, and every run states its cost before it starts.
Yes. An agent stage can name the tools a person must approve. On a hosted run the request is posted to a Slack channel you choose, and the reply decides it.
GTM Skills is free, open source, and MIT licensed. Your AI provider, Vercel, and Neon may charge for usage within the limits you choose.