RESEARCH.md
Research that is consistent between people and between runs, with the evidence line drawn in the same place every time. The method, written down: read the open roles before the About section, treat industry tags as unreliable, never infer a reporting line from a title without saying so. Attached to a Project, it makes every research output comparable with the last one.
# Research method — LinkedIn
## Reading order
1. **Open roles.** The most honest document a company publishes. They name tools, describe broken processes, and show which teams are growing. Read these before anything else.
2. **Company page posts, last 90 days.** What the company says it is doing, in its own words.
3. **The person's current role description.** What they actually own, which is narrower than the title implies.
4. **Their activity — posts and comments, last 60 days.** The single best source of a first-message reference, because it is something they chose to say publicly.
5. **Their About section.** Written once, often years ago. Useful for background, weak as a trigger.
6. **Tenure and previous roles.** Useful for judging scope and for spotting a recent move.
## What each surface reliably tells you
| Surface | Reliable for | Not reliable for |
|---|---|---|
| Open roles | tools in use, growth areas, pain | headcount, budget |
| Job title | function | scope, authority — both depend on company size |
| Seniority filter | rough level | anything at small companies, where it is frequently wrong |
| Industry tag | broad category | anything precise; it is self-reported |
| Headcount band | order of magnitude | exact size |
| Posts | current attention | long-term priorities |
| "Changed jobs" signals | a real event | why they moved |
## What LinkedIn does not tell you
Revenue, budget, tech stack beyond what a job ad names, renewal dates, whether the problem is actually felt, and who really decides. All of these are call questions, not research findings. Never present them as research.
## The evidence rule
Every claim in a brief is one of:
- **Read** — it appears in pasted text. Quote it.
- **Inferred** — reasonable, but not stated. Label it.
- **Unknown** — say so.
An assistant that blurs these produces confident briefs that get sellers embarrassed on calls.
## Scope over title
A "Head of Finance" at a 60-person company signs for software. At 6,000 people, they approve a line item someone else chose. Always read the title against the headcount band before deciding who to contact.
## What matters for us specifically
<- The number of operating entities. Visible on the company page or in the job ads, occasionally in the About section.
- Whether they have a Group Financial Controller. If they do, that is the user; if they do not, the Finance Director is doing the work personally and the message changes.>
## Stop rule
Five minutes per prospect at the qualification stage, fifteen for an account that matters. Past that you are researching instead of selling, and the extra findings rarely change the first message.- Claude Project knowledge
- ChatGPT Project files
- Agent folders
- The "what matters for us" section. The method itself needs no editing.
Everything in <angle brackets> is example content, not instructions. Replace it. The file ships filled in rather than empty on purpose: an empty template gets abandoned, a filled one gets edited.
The reference behind this file
Most of the reading order above can be automated into alerts on Advanced — that is essentially what the tier is for.
Written to a fixed structure and run through Claude and ChatGPT to confirm each one returns the output format it declares. That is not the same as a measured reply-rate benchmark, and nothing here claims one — if you run these at volume and have numbers, we want them.
Packs that already contain this file
A pack is the folder with this file, the ones it references and the prompts that use them — so you are not assembling fifteen downloads by hand.
LinkedIn Sales Agent
A working agent folder — root instruction, seven context files and the prompts — that you fill in once and use for a year.
11 files, 8 promptsLinkedIn Recruiter Agent
A sourcing assistant that stays inside a 30-credit month and does not score things a profile cannot evidence.
7 files, 5 promptsLinkedIn Job Search Agent
Every message, application and profile rewrite argues the same case about you.
4 files, 4 promptsLinkedIn AI Starter Pack
The whole toolkit as one folder, set up in twenty minutes, whatever your role is.
12 files, 12 promptsLinkedIn MCP Setup Kit
A folder you can wire an assistant to LinkedIn from, with the exposure of each route stated before the first run rather than after it.
4 files, 3 promptsOne file is a start. The folder is the point.
Files reference each other — CLAUDE.md declares the reading order, the prompts expect the files by name. The packs ship them together, already wired up.