Skip to main content
Use these prompts to turn issuer disclosures into a reviewable underwriting or claims-research packet. They surface what an issuer filed, when it filed it, and where to verify it; they do not score risk, establish coverage, set terms, or predict a claim.
For a renewal book, batch no more than 50 tickers per company call. Retain the accession number, filing date, section, and minimal supporting excerpt with every alert.

Before you run a prompt

Connect a supported client through MCP installation, then give it an API key from the environment or secret store that actually runs the client. Do not paste a key into the prompt. Start with one issuer and one prompt. A useful first result names the form, filing date, accession number, section or fact reference, source URL, and request ID when returned. These prompts create underwriting and claims-research packets; they do not score risk or set coverage terms. Keep the cited filing record with each alert, flag missing evidence, and stop on auth, validation, billing, quota, or capability errors.

Working rules

  • filings.search and sections.search do not apply an MCP date window. Retrieve candidates, then filter their returned filingDate before reporting results.
  • Treat a missing match as an incomplete search result, not proof that an issuer had no event.
  • Keep issuer disclosure, model output, and underwriting judgment in separate fields.

1. D&O renewal profile

Prompt id: insurance-d-and-o-underwriting-profile

2. Auditor-change review

Prompt id: insurance-auditor-change-keyword-scan

3. Internal-control disclosure sweep

Prompt id: insurance-material-weakness-semantic-scan

4. Restatement disclosure review

Prompt id: insurance-restatement-detection-renewal-book

5. Executive-transition alert

Prompt id: insurance-officer-departure-claims-trigger

6. Peer context for renewal discussion

Prompt id: insurance-peer-benchmark-renewal-pricing

Evidence standard

Use the cited filing as the record. Institutional holdings are delayed reports, Form 4 transactions need their transaction-code context, and any analytics output is supporting context rather than an underwriting conclusion.