Does copying install Kaspr?
No. Copying changes the clipboard. You still run the command, inspect its prompts, configure approved credentials, and review the first result.
LinkedIn prospecting in your agent
From profile brief to reviewable prospect
Kaspr helps an AI agent organize company identity, role relevance, public context, and unresolved fields without treating a returned profile as permission to contact it.
Four-state runbook
Copy, run, configure, then inspect
- 01 Copy
npx -y @okki-global/okki-go-taroballSuccess: the exact command is on your clipboard. Common error: copying a prompt marker. Next: open a supported terminal.
- 02 Run
Before: inspect the package name and runtime prompt. Success: the command completes without a package error. Common error: unsupported runtime or blocked network access. Next: review requested connections.
- 03 Configure
Before: create least-privilege test credentials. Success: the agent can call only the approved source. Common error: putting a key in a prompt or log. Next: use a known test cohort.
- 04 First result
Before: define accepted identities, functions, geography, and exclusions. Success: evidence and unknowns remain visible. Common error: confusing profile discovery with consent or intent. Next: record a human decision.
Auditable preview
One task, one record, visible reasons
Prospect result
- Company identity: review required
- Function: Revenue Operations
- Profile recency: observed date shown
- Email status: not assumed
4distinct install states
3identity checks
1human approval gate
0guaranteed replies
FAQ
Know what the workflow does not prove
Does a LinkedIn profile prove buying intent?
No. A profile or public change is context, not consent, urgency, budget, or fit. A reviewer must assess relevance.
Can the agent send automatically?
Research, drafting, approval, and sending should remain separate. Apply suppression, opt-out, regional, platform, and sender-reputation controls.
Where should API keys live?
Use a runtime secret store with minimum scope. Never put credentials in prompts, repositories, exports, screenshots, or shared logs.
How should accuracy be tested?
Use a known cohort with positive matches, exclusions, ambiguous identities, and stale records. Record false matches and unresolved fields separately.
What is a successful first result?
A record whose identity, role rationale, observation date, source context, exclusions, and unknown states can be inspected before a person approves use.