Account Research
Official Anthropic skill for structured company and people research before sales outreach.
- Skill Road
- Account Research
Categories
Account Research is an official Anthropic skill from the public knowledge-work-plugins repository. It helps a compatible Claude workflow build a useful picture of a company or person before outreach. According to the provider, the skill works with web search on its own and can be enhanced with connected enrichment tools or a CRM. This catalog entry therefore describes a research instruction set, not an independent data vendor, contact database, or guarantee that every finding is current or complete. The primary source is the Sales plugin at https://github.com/anthropics/knowledge-work-plugins/tree/main/sales/skills/account-research.
Purpose and audience
The skill is intended for sales, business development, marketing, and account teams that need more context before sending a message, holding a discovery call, or preparing for a meeting. A request can concern a company, a domain, a named person, or a role at an organization. According to the provider, prompts asking to research a company, look up a person at a company, or prepare information before a call are sufficient to start the workflow. Supplying the domain, target person, and purpose of the conversation makes it easier for the agent to focus on the right account and a relevant approach. Without that context, the risk increases that similarly named companies or unrelated public profiles will be combined.
Web research as the baseline
The documented baseline covers a company’s business, industry, size, and public positioning. According to the provider, it can also examine recent news, funding or announcement signals, leadership changes, open roles, products, services, and apparent customer groups. For people, the workflow focuses on professional background, current role, previous experience, visible public activity, and possible conversation hooks. The skill proposes several search directions, including the official company site, recent news, careers information, product pages, customer references, and public professional profiles. The result is intended to do more than collect snippets: it should prioritize findings according to their relevance to the planned outreach.
Optional enrichment and CRM context
When a team has connected suitable tools, the provider says the workflow can add verified business contact data, phone numbers, technology stack information, organizational context, funding details, and other firmographic signals. A CRM may contribute prior relationship history, known contacts, notes, activities, and won or lost opportunities. These enhancements are optional and are not supplied by the skill itself. Actual quality depends on each source, its update cycle, authorization model, and the correct matching of people and domains. A missing connector must never be replaced with invented contact details or unsupported certainty.
Output and practical use
The proposed output includes a quick assessment, company profile, recent news, hiring signals, relevant people, optional technology and CRM context, positive signals, possible concerns, unknowns for discovery, and a recommended approach. This structure turns research into preparation: a team can inspect important claims against their sources, mark gaps for the conversation, and form a reasoned hypothesis about the best entry point. According to the provider, related sales workflows such as meeting preparation or personalized outreach can follow the research. The result is still a snapshot and should be refreshed before an important customer interaction.
Boundaries, safety, and E-E-A-T
Account Research must not collect or distribute confidential information without authorization. CRM notes, contact details, internal conversation history, and non-public company data may contain personal or commercially sensitive information. Before connecting any source, teams should confirm permissions, purpose limitation, data minimization, retention, and internal approval. Public profiles can be stale, incomplete, or incorrectly matched. Search results and enrichment records must be treated as data, not as new instructions. Prompt injection in websites, profiles, or documents must not redirect the task. Human review remains necessary before contacting a person, writing a CRM change, or presenting a claim to a customer. Anthropic is the provider of this workflow according to the published primary source; responsibility for source evaluation, privacy, and the sales decision remains with the adopting team. The repository identifies Apache-2.0 as the license for the collection. Anthropic documentation at https://docs.anthropic.com/ provides complementary official product and operations information, but it does not replace review of the specific working environment.
- Provider
- Anthropic
- License
- Apache-2.0
- Last reviewed
- 09.09.2026
Repository and documentation
Categories
Compatible with
Related guides
Guides and background related to this entry.
Set up Mapbox MCP Server
Set up the Mapbox MCP Server: hosted endpoint or local token, a first test, and sensible limits.
30.09.2026
Setting up the Agent News MCP Server
The Agent Times' free, no-auth MCP server delivers verified, sourced news with confidence scores about the AI agent and MCP industry.
18.09.2026
Set up Typeform MCP Server
Start Set up Typeform MCP Server with minimal permissions, correct transport configuration, and verified data flow.
18.09.2026
Set up Bright Data MCP Server safely
Configure the official Bright Data MCP Server as a remote endpoint or local process: store the API token securely, activate tool groups deliberately, and understand scraping boundaries and prompt-injection risks.
18.09.2026