Skip to main content
These prompts describe useful workflows. Results depend on the records available in your authorized workspace; a company named here is not guaranteed to be present.

Start with your sources

List my Signal Agents and the Sheets they are linked to. Show each agent’s schedule and status.
Use the returned Sheet IDs to narrow later searches rather than relying on a display name.

Find an Account and its Leads

Find Accounts with the domain stripe.com. Use the matching Account IDs to find associated Leads, and show their role, saved AI score, and linked findings from the selected evaluation.
The API Quickstart uses an illustrative Stripe hiring scenario and a fictional Lead. The Lead score of 82 and Account score of 64 illustrate separate assessments.

Prioritize recent results

Show up to 10 Leads with an AI score of at least 80 whose latest saved signal falls within November 2026. Include the source Sheet and evidence behind each result.
Ask for explicit date boundaries when time matters. A record’s latest_signal_at can refer to a different finding from its selected evaluation.

Review a Lead’s history

For this Lead ID, show the saved findings newest first. Compare the available evidence and each finding’s qualification. Keep missing scores as missing rather than using the current Lead score.
History includes retained findings rather than an immutable log of every scheduled check. The same configured source can have multiple findings while contributing only once to signal_count.

Explain a finding

Explain why this Account was discovered. Separate the recorded qualification and AI score from your interpretation, and cite the returned source links.
The assistant can summarize saved results. It should not describe a profile match or social interaction as proven purchase intent, invent an observation date, or claim to have rerun the agent.

Review other evaluations

Show the saved evaluations for this Lead. For each decision, list its linked findings and compare their evidence. Continue any truncated finding previews using that evaluation’s ID. Keep evaluation scores separate from finding scores.
Use get_lead_evaluations or get_account_evaluations for decisions, and the corresponding signal-history tool with evaluation_id for linked-finding continuation.