Prospect research prompt for Claude and ChatGPT
Paste a company name, get a structured dossier you can act on instead of a wall of text.
Research the company named below and produce a prospect dossier. Rules: - Output exactly the format below, including the marker comment on the first line. - Never invent. Leave any field blank if you cannot verify it from a real source. - Do not guess emails, phone numbers, or names. Blank is always better than plausible. - Keep Observations to what the sources actually say. Company: <company name> <!-- anguardia-dossier v1 --> # Dossier: <Company Name> ## Company - Industry: <industry> - Website: <url> - Location: <city or region> - Source: <cold | referral | inbound | research> ## People | Name | Role | Email | Phone | LinkedIn | |------|------|-------|-------|----------| ## Observations <free markdown: pain points, current software, manual workflows, the personalised observation for outreach> ## Suggested tasks - [ ] <task title> | due: <YYYY-MM-DD, optional> ## Suggested outreach <the first message, written for the decision maker, under 150 words>
What does this prompt do?
It asks Claude or ChatGPT to research one company and return a Dossier v1 document — fixed sections for company facts, people, observations, suggested tasks, and a short outreach draft.
The first line is a marker comment (<!-- anguardia-dossier v1 -->). That marker tells a machine the structure is intentional, not decorative.
What do you get back?
Five sections, always in the same order. Empty fields stay blank — the prompt forbids inventing contacts.
- Company
- Industry, website, location, and source (
cold | referral | inbound | research). Unknown bullets are ignored by importers — never invented into fields. - People
- A markdown table: Name, Role, Email, Phone, LinkedIn. Empty cells are null. The outreach draft’s recipient is always the first person in the table.
- Observations
- Free markdown: pain points, current software, manual workflows, and the personalised observation for outreach. Knowledge that does not fit a field stays here — it is never discarded.
- Suggested tasks
- Checkbox lines, optionally with
| due: YYYY-MM-DD. These become follow-ups you can date and act on. - Suggested outreach
- The first message for the decision maker, under 150 words — written during research, not by a CRM later.
Worked example
Fabricated company only — not a real prospect.
<!-- anguardia-dossier v1 --> # Dossier: Harborline Freight ## Company - Industry: Logistics - Website: https://harborline.example - Location: Singapore - Source: research ## People | Name | Role | Email | Phone | LinkedIn | |------|------|-------|-------|----------| | Maya Chen | Head of Operations | maya@harborline.example | | https://linkedin.com/in/maya-chen-example | | Sam Ortiz | IT Manager | | | | ## Observations Harborline still coordinates dock schedules in shared spreadsheets. Ops leads re-enter the same shipment status into three systems. Recent job posts mention "warehouse systems modernisation" but no vendor named publicly. ## Suggested tasks - [ ] Send intro email to Maya | due: 2026-08-05 - [ ] Sketch a 15-minute discovery agenda - [ ] Check LinkedIn for mutual connections ## Suggested outreach Hi Maya — saw Harborline is still running dock schedules in spreadsheets while posting about warehouse systems modernisation. I help logistics ops teams replace that re-entry loop with one status view the floor and customers can trust. Open to a short call next week?
Why a fixed format instead of a normal research reply?
Fixed headings mean a machine can read the dossier, not just a human. Unstructured research is a chat transcript; Dossier v1 is a contract.
The marker <!-- anguardia-dossier v1 --> (or the legacy <!-- founder-os-dossier v1 -->) identifies the format. Format changes bump the version to v2; parsers support both during transition.
| Criterion | Unstructured research | Dossier v1 |
|---|---|---|
| Can a machine parse it? | No — free prose | Yes — fixed headings |
| Contacts as data? | Buried in paragraphs | People table rows |
| Dated next actions? | Maybe, if you dig | Suggested tasks with due dates |
| Outreach recipient? | Unclear | First person in the table |
Do you need Anguardia to use it?
No. The prompt is free and useful on its own. Anguardia turns the same output into a company, contacts, dated follow-ups, and an outreach draft automatically — without retyping — when you want that next step.
Frequently asked questions
- Does this work with ChatGPT?
- Yes. Paste the prompt into ChatGPT or Claude, put the company name where indicated, and ask it to research. Both models can follow the fixed headings.
- Do I need Anguardia to use it?
- No. The prompt is free and useful on its own — you get a structured dossier you can act on. Anguardia is optional: it turns that same output into a company, contacts, dated follow-ups, and an outreach draft without retyping.
- Can I change the format?
- You can edit the prompt however you like. The Anguardia importer only parses Dossier v1 (or the legacy founder-os marker). Change the headings or drop the marker and Anguardia cannot read the output.
- Will it invent contact details?
- The prompt tells the model not to invent — leave any field blank if it cannot verify a real source. Blank is always better than plausible. Still verify emails and phones before you send anything.