How is this different from a generic onboarding checklist?

A generic checklist teaches a textbook FAQ; this builds the curriculum from your real assets and your real queue. The cheat sheet maps the top 20 issues a new hire will actually face — pulled by volume from your ticket export — to your own articles, macros, and escalation lines, so they learn the standard your team already uses instead of a generic one.

Why practice tickets instead of just having the new hire read the docs?

Because reading is passive and the voice is learned by writing. Practice tickets make the new hire draft a reply, then compare it to a model answer in your voice — active recall, not highlighting — and a couple of the cases are deliberately ones that should be escalated, so they practice judgment before they touch a live customer.

Does this replace shadowing a senior agent?

No. Claude builds the curriculum, but it doesn't replace the mentor. The ramp plan is deliberately graduated — read, then shadow a human, then send supervised replies a senior approves, then go solo on easy tiers — because edge cases and the feel of a hard conversation are learned beside a person, not from a doc.

Can I reuse the onboarding doc for the next hire?

Yes — that's the point of assembling it into one onboarding.md. Re-run the build with a fresh tickets.csv so the top-20 issue list stays current, and the ramp gets sharper every hire instead of staler. Just keep the source docs current too, or every new hire inherits last quarter's answers.

Ramp a new support hire in their first week

when to reach for this

A new support agent usually learns by osmosis — shadowing, asking, and getting things wrong on live customers for a month before they sound like the rest of the team. The fix is to stop ramping from scratch every hire: you already own the assets that define how your team answers — the voice and policy doc, the macro library, the help center, and the real top-issue clusters from your tickets. This turns those into a structured first-week ramp — a personalized onboarding guide, a "top 20 issues and how we answer them" cheat sheet built from real data, practice tickets with model answers, and a graduated shadow→supervised→solo plan with checkpoints — so a new hire is useful in days, and consistent from their first reply because they learned on the same source-of-truth docs everyone else uses.

gather this first

Your existing source-of-truth docs: the support-voice.md (or policy doc), the macros.md canned-response library, and your help-center article list — the same docs the team already answers from, so the new hire learns the real standard, not a generic one.
A recent ticket export as tickets.csv — drop it into the chat in Claude Desktop, scrubbed of names, emails, and account numbers to [customer] first — so the ramp teaches the issues a new hire will actually face, not a textbook FAQ.
Your escalation rules and tier map: which issues a new agent can solve solo, which need a senior, and where the line is — so the cheat sheet and the ramp plan know what "easy tier" actually means for your team.

the workflow

Inventory the assets and pull the real top issues

Start by pointing Claude at the docs you already have and the recent ticket export, so the ramp is built from how your team actually works and the queue a new hire will actually see — not a generic onboarding template.

you ask

Here are our existing docs — support-voice.md, macros.md, and our help-center article list — plus a recent tickets.csv (names scrubbed to [customer]). First, confirm what you can see: the row count and date range of the tickets, and a one-line summary of each doc. Then group the tickets by underlying issue (not wording) and give me the top 20 issues by volume with counts, flagging which ones a brand-new agent could handle solo versus which need a senior. Don't write any training material yet — just the inventory and the ranked issue list.

what you get back A read-back of your assets plus a ranked, count-backed list of the top 20 real issues — "reset password (47), where's my order (38), refund request (29, senior)…" — with the solo-vs-senior flag, so the ramp teaches the actual queue rather than a guess.

Verify the issue counts against the raw export here — every downstream piece (cheat sheet, practice set, ramp plan) keys off this list, so a wrong ranking trains the new hire on the wrong issues.
Build the "top 20 issues" cheat sheet

Turn that ranked list into the single page a new hire keeps open beside the queue. For each issue: what it is, where it's documented, the macro that fits, and the escalation line — so they can find the right answer in seconds instead of asking the room.

you ask

Using the top 20 issues and our docs, build a cheat sheet. For each issue give me one row: the issue in plain words, the help-center article that documents it (by title), the macro from macros.md that fits (by name), one line on how to recognize it, and the escalation line — when to stop and hand it to a senior. Where no article or macro exists yet, mark it [no article yet] / [no macro yet] rather than inventing one. Keep each row scannable.

what you get back A one-page cheat sheet, 20 rows, each mapping a real issue to its article, its macro, a recognition cue, and a clear escalation line — with [no article yet] / [no macro yet] markers exposing the gaps in your docs instead of papering over them.

The [no article yet] markers are a bonus output — they show a senior exactly where the docs have holes, which is its own to-do list for the help center and macro library playbooks.
Generate practice tickets with model answers

Reading is passive; the fastest way to learn the voice is to write a reply and compare it to a good one. Build practice tickets from real, anonymized past cases so the new hire drafts first, then checks against the model — active recall, not highlighting.

you ask

From the anonymized tickets, create 10 practice tickets across our top issues and a range of difficulty (mostly easy-tier, a couple that should be escalated). For each: the incoming customer message (with any names or details already as [customer], [order number], etc.), then a separate 'model answer' written in our voice using the right macro — and for the escalation ones, the model answer is 'recognize this is out of scope and hand it to a senior, here's the line to use.' Format so the new hire can read the ticket, draft their own reply, and only then reveal the model answer.

what you get back Ten practice tickets that mirror your real queue, each paired with a hidden model answer in your voice — including a couple where the right move is to escalate, not solve — so the new hire practices judgment, not just typing, before they touch a live customer.

Make sure the practice cases are drawn from anonymized past tickets, never live ones — the new hire is learning, and a real customer's details don't belong in a training doc.
Write the week-by-week ramp plan

Sequence the week so confidence is earned, not assumed: read the docs, then shadow a senior, then send supervised replies, then go solo on easy tiers — with a checkpoint at each gate so a manager knows the new hire is ready before the stakes go up.

you ask

Write a 5-day ramp plan for a new support hire using everything above. Day-by-day: which docs to read and the cheat sheet to absorb, then shadowing a senior on live tickets, then drafting supervised replies a senior approves before sending, then handling easy-tier tickets solo with a senior reviewing afterward. At each gate include a concrete checkpoint — what the new hire should be able to do before moving to the next stage — and which practice tickets to do when. Keep it realistic for someone learning, not a sprint.

what you get back A graduated day-by-day plan — read → shadow → supervised → solo on easy tiers — with a clear, checkable gate at each stage, so the new hire never gets handed more responsibility than they've demonstrated, and a manager can see exactly where they are.
Assemble it all into one `onboarding.md`

Pull the inventory, the cheat sheet, the practice set, and the ramp plan into a single handoff doc a manager can hand over on day one — and reuse for the next hire by swapping in a fresh ticket export.

you ask

Assemble everything into one onboarding.md: a short welcome and how-to-use intro, the top-20 cheat sheet, the 10 practice tickets (with model answers in a clearly separated section so they're not spoiled), and the 5-day ramp plan with its checkpoints. Add a one-line note at the top for the manager: 'to reuse this for the next hire, re-run with a fresh tickets.csv so the issue list stays current.' Keep it skimmable with clear headings.

what you get back A single paste-ready onboarding.md — welcome, cheat sheet, practice set, and ramp plan in one place — that a manager hands over on day one and regenerates for the next hire from a fresh export, so the ramp gets sharper every time, not staler.

Save this file. It's the artifact you reuse — and once your new hires are live, their early replies are exactly what the QA the queue and coach the team to one voice playbook reviews.

make it your own

Build the assets first: the ramp is only as good as the docs it's built from. If you don't have them yet, run Set the support voice and policy every reply inherits, Build a canned-response library you'll actually reuse, and Build a help center from the questions you actually get first — then this playbook turns them into a curriculum.
Source the issue list from real data: run Find what the queue is really about on your export before step 1 so the top-20 cheat sheet maps to actual demand, and the new hire learns the issues that genuinely flood the queue.
QA the graduate: once the new hire is sending supervised and then solo replies, run QA the queue and coach the team to one voice on their early tickets — the ramp teaches the standard, and QA confirms it landed.
Keep the cheat sheet auto-fresh (Power Track): set a /ramp-refresh custom command or a scheduled agent (see the Playbook's Features tab) to rebuild the top-20 cheat sheet from each month's top clusters, so the next hire always trains on the current queue. Custom commands and scheduled agents are the opt-in Power Track — on Desktop you re-run the step-1 prompt by hand with a fresh tickets.csv whenever you onboard someone new.

watch out for

Build the practice tickets from anonymized past cases, never live ones — scrub names, emails, and account numbers to [customer] before the export goes into the chat, because a training doc is the last place a real customer's details should end up.
The cheat sheet is a draft, not gospel — a senior agent must verify every issue-to-macro mapping, every escalation line, and especially every [no article yet] gap before a new hire trusts it, because a confident wrong answer on the cheat sheet becomes the new hire's default.
A new hire still shadows a human — Claude builds the curriculum, but it doesn't replace the mentor. The supervised and shadow stages exist precisely because judgment, edge cases, and the feel of a hard conversation are learned beside a person, not from a doc.
Keep the source docs current or the ramp teaches stale answers — if macros.md, the help center, or the policy doc are out of date, every new hire inherits the old answer from day one. Regenerate the ramp from fresh docs and a fresh export, not last quarter's.

you'll end up with A reusable `onboarding.md` built from how your team actually works — a top-20 cheat sheet mapping real issues to your articles, macros, and escalation lines; practice tickets with model answers; and a shadow→supervised→solo ramp plan with checkpoints — so a new hire is useful in days instead of weeks, and consistent from their first reply because they learned on the same source-of-truth docs as everyone else.

Questions people ask

How is this different from a generic onboarding checklist?: A generic checklist teaches a textbook FAQ; this builds the curriculum from your real assets and your real queue. The cheat sheet maps the top 20 issues a new hire will actually face — pulled by volume from your ticket export — to your own articles, macros, and escalation lines, so they learn the standard your team already uses instead of a generic one.
Why practice tickets instead of just having the new hire read the docs?: Because reading is passive and the voice is learned by writing. Practice tickets make the new hire draft a reply, then compare it to a model answer in your voice — active recall, not highlighting — and a couple of the cases are deliberately ones that should be escalated, so they practice judgment before they touch a live customer.
Does this replace shadowing a senior agent?: No. Claude builds the curriculum, but it doesn't replace the mentor. The ramp plan is deliberately graduated — read, then shadow a human, then send supervised replies a senior approves, then go solo on easy tiers — because edge cases and the feel of a hard conversation are learned beside a person, not from a doc.
Can I reuse the onboarding doc for the next hire?: Yes — that's the point of assembling it into one `onboarding.md`. Re-run the build with a fresh `tickets.csv` so the top-20 issue list stays current, and the ramp gets sharper every hire instead of staler. Just keep the source docs current too, or every new hire inherits last quarter's answers.