OpenClaw for GTM: Give Your Agent Prospecting, Enrichment and Outbound Skills

OpenClaw can already do go-to-market work. It just has nothing to do it with. It runs on your own machine, answers you in Slack, Telegram, WhatsApp or iMessage, schedules its own jobs and loads skills that follow the AgentSkills standard. Ask it to find fifty heads of RevOps at Series B fintechs and it will open a browser, search, and hand you a list that is half guesses, because it has no B2B data, no email verification and no account at any provider that has either.

That is the gap Cargo’s skills fill. Install them and the same request runs against licensed data providers through one CLI, with verified emails, a cost receipt and a record of the run. The skills are on ClawHub under @cargo-ai, they work the same way in Hermes Agent, and a new Cargo account starts with 100 free credits and no card.

What the skills give your agent #

Two packs, both public and both published to ClawHub on every change.

The platform pack (19 skills). One skill per Cargo domain: cargo-gtm for sourcing, enrichment, scoring and outreach drafts, cargo-storage for querying your data, cargo-orchestration for running workflows and batches, cargo-mailbox-management for sending, cargo-diagnostics for explaining a run that went wrong, and a router skill, cargo, that tells the agent which of the others to load.

The job pack (37 skills). Each named for the job rather than for Cargo, so the agent finds it by what you asked: find-work-email, build-tam-list, find-stakeholders, track-job-changes, track-funding-rounds, find-companies-using-tech, verify-email-list, score-leads, research-account. The skills carry the procedure, not the data. The data comes from the providers Cargo connects to, which is why every skill needs the cargo-ai CLI and a signed-in workspace.

Install in OpenClaw #

Install the CLI first. OpenClaw checks each skill’s requirements, and a Cargo skill without cargo-ai on the path stays installed but hidden: openclaw skills info cargo-gtm reports Needs setup and Visible to model: no until the binary exists.

npm install -g @cargo-ai/cli

openclaw skills install @cargo-ai/cargo
openclaw skills install @cargo-ai/cargo-gtm
openclaw skills install @cargo-ai/find-work-email

That installs into the current agent’s workspace. Add --global to share the skills across agents. To take the whole platform pack in one command instead of one skill at a time:

npx skills add getcargohq/cargo-skills --agent openclaw -g

Check that the agent can see them:

openclaw skills info cargo-gtm   # ✓ Ready, Visible to model: yes

OpenClaw’s own guidance applies to these as to any third-party skill: read it before you enable it. Every skill is plain markdown, and the source is in getcargohq/cargo-skills and getcargohq/gtm-skills.

Sign the agent in #

Sign-up runs entirely in the terminal and creates the account on first use, so there is no separate sign-up step:

cargo-ai login --email you@company.com                 # sends a code, then exits
cargo-ai login --email you@company.com --code 123456
cargo-ai whoami

Do this on the machine that runs the OpenClaw gateway, because that is where the skills execute. On a box that runs unattended, use an API token instead of your personal session: create one under Settings → API, scoped to one workspace, and give the gateway CARGO_API_TOKEN. A saved login acts as you across every workspace you belong to. A token can be scoped down and revoked from the dashboard.

The 100 free credits come to roughly 5,000 leads sourced, 1,000 verified-email enrichments, or 50 fully enriched contacts.

What to ask it #

Ask the way you would ask a colleague, from whichever channel the agent lives in. The skills’ triggers are written in that voice:

• “Find 50 heads of RevOps at Series B fintechs in France.”
• “Find work emails for the people in this CSV.”
• “Who on our customer list changed jobs this month?”
• “Which companies in this list use Snowflake?”
• “Research Acme before my call and give me a brief with sources.”

The agent loads the router, picks the skill, and runs the job through the CLI, and each run reports what it spent.

The skills also carry the rules for anything that touches a person. Every step that contacts someone is gated on a consent basis, a suppression list and a volume limit, and purchased or scraped lists and consumer targeting are refused. An agent with Cargo’s skills drafts outreach for your sequencer or for a Cargo mailbox. It does not bulk-mail strangers because someone asked it to in a group chat.

Put it on a schedule #

OpenClaw’s built-in scheduler turns a request into a standing one. A list every Monday morning, delivered to a Slack channel:

openclaw automations create "0 8 * * 1" \
  "Using the Cargo skills, find companies in our ICP that raised a round in the last 7 days, find the head of sales at each, and post the list with verified emails." \
  --name "Monday funded accounts" \
  --tz "Europe/Paris" \
  --session isolated \
  --announce --channel slack --to "channel:C1234567890"

This is the right shape for a job one person wants, delivered to them, where a missed week costs nothing.

When the job should leave the agent #

A personal agent runs on one person’s machine, with that person’s session, until the laptop closes. That is fine for the Monday list. It stops being fine when the job has to run whether or not that machine is on, when the team depends on its output, or when someone has to answer for what it did six weeks ago.

At that point the agent should stop running the job and start building it. The same CLI declares the work as a Cargo play, in TypeScript, in a repository: it runs on Cargo on its schedule or when data changes, retries on failure, and leaves a trace of every run.

cargo-ai project init my-gtm --cookbook new-hire-detection
cd my-gtm
npm install               # the cookbook's dependencies
cargo-ai project plan     # the diff against the workspace
cargo-ai project deploy   # needs a login

Ask OpenClaw to do it, with the cargo-project skill loaded, and it will. That is the division of labour that holds up: the agent drafts and changes the system, Cargo holds the credentials, the data, the schedule and the audit log, and a person approves the diff. We wrote up the operating model in run GTM with coding agents.

Hermes Agent #

Hermes Agent, from Nous Research, follows the same AgentSkills standard, so the same packs work. The skills CLI installs them into Hermes’s skill directory:

npm install -g @cargo-ai/cli
npx skills add getcargohq/cargo-skills --agent hermes-agent -g   # → ~/.hermes/skills

Sign in the same way, on the machine Hermes runs on. If you are moving from OpenClaw, hermes claw migrate imports the skills you already installed there.

Frequently asked questions #

Sources #

• OpenClaw: repository, skills, ClawHub, and skills CLI.
• Hermes Agent: repository and skills.
• Cargo: cargo-gtm on ClawHub, cargo-skills, gtm-skills, agent install instructions, and cookbooks.