ACCOUNT RESEARCH, WITH THE WORK SHOWN

6sense maps the buying committee behind the signal.

Turn an ideal customer profile brief into a reviewable company and decision-maker research trail inside your AI agent. Examine the inputs, source notes, and confidence before any outreach leaves your desk.

Inspect the evidence model
npx -y @okki-global/okki-go-taroball
AI agent research workspace showing an account list and evidence panel

RESEARCH CONTRACT

Define what counts before ranking an account.

Each run begins with explicit criteria, not a hidden score. The grouped specification keeps targeting logic visible to RevOps and sales leadership.

Ideal customer profile

Firmographic scopeMarket, employee band, business model, and operating region
Technographic contextObserved stack signals with a source note and checked date

Buying committee

Decision rolesEconomic buyer, functional owner, evaluator, and likely user
Evidence thresholdInclude, exclude, or hold for human review
4 statesCopy → Run → Configure → Result
3 layersTask → Finding → Evidence
1 reviewerHuman approval remains explicit
0 hidden claimsUnknown is a valid outcome

CAPABILITY EVIDENCE

Research tasks, not another seat-based dashboard.

01

Find companies

Input
An ICP expressed in plain language
Output
A shortlist with inclusion reasons

Limit: broad criteria create broad results; refine the brief before trusting rank order.

02

Map decision roles

Input
Target account and buying motion
Output
Role hypotheses and contact candidates

Limit: titles do not prove authority, so the agent marks role inference for review.

03

Prepare context

Input
Account, trigger, and research question
Output
Source-linked notes for a rep

Limit: intent signals indicate possible interest, not a confirmed buying project.

04

Draft outreach

Input
Approved facts and message constraints
Output
A draft held for human review

Limit: sending, consent, opt-out, and regional compliance remain your responsibility.

INSTALLATION RUNBOOK

Four distinct states. No false finish line.

  1. 01

    Copy

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

    Success signal: the exact command is in your clipboard.

  2. 02

    Run

    Execute it in a supported local agent environment. Review the package prompt and terminal output before proceeding.

  3. 03

    Configure

    Add only the data-provider credentials your workflow needs. Store API keys in the runtime's secret mechanism, never in a prompt.

  4. 04

    First result

    Start with a narrow ICP query, inspect source notes, and approve or reject the evidence before using the result.

REVIEW NOTES

Questions an evaluator should ask.

Does copying the command install the skill?

No. Copying only places text in your clipboard. You must run it, review the runtime output, configure access, and test a first result.

Is every contact automatically correct?

No. Contact freshness and role relevance vary by source and configuration. Verify critical fields before outreach.

Does the skill send cold email automatically?

The operating model keeps draft review explicit. Your organization controls sending tools, suppression, consent, opt-out, and sender authentication.

How should we judge an intent signal?

Treat it as prioritization evidence, not proof of a purchase. Check the signal definition, time window, account match, and corroborating context.

Where should API keys live?

Use the agent runtime's supported secret store or environment configuration. Do not paste credentials into chat history or committed files.

How do we remove or update it?

Use the package and runtime controls documented during installation. Revoke provider keys separately when retiring a workflow.