A Claude Skill that takes a company domain and produces a structured account brief: company snapshot, strategic priorities pulled from earnings calls or recent press, tech stack signals, key personas and their reporting lines, and three hypothesized pain points mapped to your product. Run it before every discovery call.

What you’ll need

Claude.ai or Claude Code with web fetch enabled
A company domain or LinkedIn URL as input
Your product positioning doc (1-pager) loaded into the Skill folder
Optional: a paid signal source like Apollo or BuiltWith for stack data

Setup

Drop the Skill into Claude. Save the account-research.skill directory to ~/.claude/skills/ (Claude Code) or import into a Claude.ai project. The Skill exposes research_account(domain) as the entrypoint.
Load context files. Place your one-pager, ICP rubric, and competitor list in the Skill’s references/ subdirectory. The Skill reads these on every run so the brief is grounded in your positioning, not generic AE research.
Configure output format. The default output is a Markdown brief with five sections. Edit SKILL.md if your team prefers Notion blocks, a Google Doc, or a Slack-friendly summary.
Run on a test account. claude run research_account --domain acme.com. Inspect the brief, then refine the prompts in SKILL.md until the output matches what your best AEs would write by hand.

How it works

The Skill runs four sub-tasks in sequence. First, it fetches the company homepage, About page, and careers page to build a snapshot. Second, it searches for recent earnings transcripts, press releases, or funding announcements from the last 90 days to surface strategic priorities. Third, it pulls public LinkedIn data on the buying committee — typical RevOps, Marketing Ops, and Sales Ops titles — and notes tenure and prior companies. Fourth, it synthesizes pain hypotheses by cross-referencing the snapshot against your positioning doc.

The output is deliberately opinionated. Instead of dumping facts, the Skill ranks the three most likely entry points and suggests a discovery question for each. Reps can adopt, edit, or reject — but they always start from a thesis, not a blank page.

Watch-outs

Stale public data. LinkedIn scraping breaks frequently and earnings transcripts lag a quarter. If freshness matters, layer a paid source like ZoomInfo or Apollo upstream.
Hallucinated org charts. Claude will invent reporting lines if pushed. Constrain the persona section to titles actually visible on LinkedIn or the company site, and require a citation for each name.
Positioning drift. If your one-pager is six months old, the pain hypotheses will reflect old messaging. Re-sync the references directory whenever marketing updates positioning.

Stack

Claude — orchestration, web fetch, and synthesis
Apollo or BuiltWith (optional) — tech stack and headcount signals