Agency LinkedIn Client Onboarding Prompt
A filled ICP, voice and product brief per client, plus an explicit list of what the kickoff failed to establish. Agency work fails at onboarding: the kickoff produces notes, the notes produce a doc nobody reads, and three weeks later the copy is wrong in the same way for the fourth time. This prompt converts a kickoff transcript into the four files the work actually runs on, and lists what the call did not answer.
You are an agency operations lead converting a client kickoff into working documents. You are strict about the difference between what the client said and what you assumed.
## The client
{{CLIENT}}
What they sell: {{PRODUCT}}
The engagement objective: {{OBJECTIVE}}
## Source
[PASTE THE KICKOFF TRANSCRIPT OR YOUR NOTES IN FULL]
[PASTE 3–5 EXAMPLES OF THEIR EXISTING WRITING: posts, emails, website copy]
## Task
Produce four documents, each complete enough to hand to someone who was not on the call.
1. **ICP** — firmographics in LinkedIn's own bands, buying committee, the trigger, and the exclusions. Mark every attribute as "stated by client" or "inferred", and never silently upgrade an inference.
2. **Voice** — derived from the writing samples, not from what they said about their voice. Those two things disagree more often than they agree. Include: sentence length, vocabulary they use, vocabulary they avoid, how they handle claims they cannot prove, and five real sentences from the samples as reference.
3. **Proof and claims** — what may be said, what may be said without naming the customer, and what may not be said at all. Flag anything that would need legal or client sign-off.
4. **Approval rules** — who approves what, how long they get, and what happens when they do not respond. If the call did not establish this, say so loudly: it is the single most common cause of an engagement stalling.
Then list every question the call left unanswered, ranked by how much damage each gap will do.
## Constraints
- Never fill a gap with a sensible assumption. An assumption in an onboarding doc becomes a fact by week three.
- Voice rules must quote actual sentences from the samples.
- If the samples are ghostwritten or inconsistent, say so — it changes the whole engagement.
## Output format
### ICP.md
Full file content, ready to save.
### VOICE.md
Full file content, ready to save.
### PROOF.md
Full file content, ready to save.
### APPROVALS.md
Full file content, ready to save.
### Unanswered
| Question | Why it matters | Damage if we guess | Ask by when |
|---|---|---|---|- You just finished a kickoff call and have a transcript.
- You are onboarding a client onto an AI-assisted workflow and need the context files.
- Your team keeps producing copy the client rejects for reasons nobody wrote down.
- Record the kickoff and paste the transcript. Notes lose the client’s actual phrasing, which is what the voice file is built from.
- Paste real writing samples. The voice derived from samples and the voice the client describes are usually different documents.
- Send the "unanswered" table back to the client the same day. It is the most professional-looking artefact in the whole onboarding and it protects you later.
- Save the four files into the client folder and attach them to that client’s Project.
The output schema it demands
The single highest-leverage part of any prompt. Without a declared shape you get prose you then have to reformat; with one, you get sections you can act on — and you can tell at a glance when the model skipped something.
### ICP.md (full file) ### VOICE.md (full file) ### PROOF.md (full file) ### APPROVALS.md (full file) ### Unanswered (ranked by damage)
What it looks like filled in
Illustrative, not a promise — this is the shape of a good answer rather than a guarantee of one. The scenario is the same fictional company used across the whole toolkit.
Kickoff transcript with a new client selling supply-chain visibility software, 45 minutes, plus 3 LinkedIn posts they had already published.
ICP.md, extract — firmographics as stated on the call: 500-5,000 staff, manufacturing. Trigger: not stated — flagged in the unanswered list rather than guessed.
VOICE.md, extract — from the 3 samples: short sentences, no exclamation marks, one customer number per post. The client described their own voice as "bold and energetic," which disagrees with the samples — the file follows the samples, not the client's description of themselves.
Unanswered — approval turnaround time was never discussed on the call. Flagged as high-damage and sent back to the client the same day, rather than assumed.
Attach these once and stop retyping your business
This prompt works pasted into a blank chat. It works considerably better inside a Project with these attached, because then the ICP, the product and the limits are already in context and every answer inherits them.
ICP.md
Every prompt you run afterwards knows who you sell to, in terms LinkedIn can filter on.
Free downloadBRAND-VOICE.md
Drafts arrive in your voice on the first attempt instead of after four rounds of "less corporate".
Free downloadCOMPANY.md
The assistant stops describing your company the way its training data describes companies in your category.
Free downloadThe reference behind this prompt
If you run outbound for several clients, seat structure and cost per seat is the first thing the engagement maths depends on.
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.
Take the whole folder instead
This prompt, the files it expects and the rest of the sequence, as a zip you can drop into a Claude Project, a Cursor workspace or an agent repo. Free, no account.