Brand Logo

Agent-native prospecting · evidence first

Salesloft turns a market question into a prospect list you can interrogate

Give your AI agent an ICP, a territory, and the evidence standard. The skill returns structured research with source context and review states, without forcing another seat-based workspace into your stack.

Review the runbook

Copying is step one; you still run and configure the skill in your agent.

INSTALLnpx -y @okki-global/okki-go-taroball
SAMPLE TASK
Find B2B infrastructure teams hiring RevOps, then show the evidence behind each match.
AI agent prospecting research interface with evidence columns

Grouped specification ledger

Define the operating contract before evaluating an output

The values below are evaluation requirements, not performance promises. Runtime behavior, provider coverage, and downstream retention must be checked in the environment where the skill runs.

Research inputs

ICP definition
Firmographic, technographic, territory, exclusion, and confidence criteria
Signal policy
Named evidence types, retrieval date, and minimum corroboration required

Review controls

Record state
Verified, config-dependent, not disclosed, or requires human review
Outreach boundary
No message advances without relevance, suppression, and regional compliance review

Environment checks

Credentials
Provider keys remain in the runtime secret store with narrowly scoped permissions
Deliverability
SPF, DKIM, DMARC, mailbox health, and bounce handling stay outside any prospect-data claim

Vertical verification timeline

A result becomes useful through four review gates

01

Frame the market question

Record buyer role, company profile, territory, exclusions, and the business event that would count as a meaningful signal.

02

Collect source context

Attach public evidence, retrieval timing, and provider-specific caveats so an analyst can distinguish an observed fact from an inferred match.

03

Resolve uncertain records

Use waterfall enrichment only where additional sources add value; preserve disagreement instead of silently treating a single source as ground truth.

04

Approve downstream action

Check suppression, sender reputation, message relevance, and applicable regional obligations before export, drafting, or any human-authorized send.

Six evidence controls

Inspect the method at the same resolution as the result

01

ICP schema

Separate required attributes, preferred signals, exclusions, and unknown fields.

02

Source trail

Carry URLs and retrieval context beside each structured prospect record.

03

Confidence state

Label inference and provider disagreement for deliberate human resolution.

04

Waterfall restraint

Query another source only when the expected evidence justifies cost and latency.

05

Review checkpoint

Keep audience choice, message approval, and send authority with a person.

06

Export discipline

Apply retention, suppression, and access policies before records leave the agent.

Four-state runbook

Copy is not install; install is not a first result

  1. 01

    Copy

    npx -y @okki-global/okki-go-taroball

    Success signal: the exact command is in your clipboard.

  2. 02

    Run

    Paste the command into the terminal used by your supported AI-agent environment. Read the package prompt before allowing execution.

    Success signal: the installer completes without a package or permission error.

  3. 03

    Configure

    Add only the provider credentials your chosen workflow requires. Keep keys in the runtime secret store; never paste them into prospecting prompts.

    Success signal: the agent can enumerate the installed skill without exposing secrets.

  4. 04

    First result

    Begin with a narrow ICP and request evidence fields. Review sources, uncertain matches, and exclusions before exporting or drafting outreach.

    Success signal: a structured sample appears with reviewable provenance.

Install FAQ

Questions to settle before the first run

What does the npm command do?

It invokes the published installer through npx. Review terminal output, requested permissions, package provenance, and runtime policy before continuing.

Does copying the command install anything?

No. Copy only places the exact text on your clipboard. You decide when and where to run it.

Will the skill send cold email automatically?

The recommended workflow keeps targeting, message approval, suppression, and send decisions under human control. Follow CAN-SPAM, CASL, GDPR, and other rules applicable to recipients and regions.

How should API keys be handled?

Use the agent runtime’s secret management mechanism, scope permissions narrowly, rotate credentials, and confirm deletion behavior for stored results.