Every role-specific line in the onboarding plan is still in brackets
Your new hire starts Monday. Their first-month plan is still a page of blanks. Claude Docs can fill those blanks from four files your team already keeps, and name the file behind every line. A hiring manager writes one of these plans for every new hire and every internal move. The plan says what to read, who to meet, and the goals for day thirty, day sixty and day ninety, meaning what the new hire should have done by each of those days. The team here is Customer Insights at a company called Tallowmere Goods. I made up the company, the four files, and every person in them. The new hire is Dana Whitcombe, an analyst, and her manager is Ines Carvalho.
Anthropic’s template repeats “[Goal aligned to role]” three times
Don’t overload Day 1
Deep work starts Week 2.
This template comes from Anthropic itself. Anthropic publishes an onboarding skill for Claude, and a skill is a written set of instructions Claude follows for one kind of job. The onboarding skill wants five facts: the new hire's name, role, team, start date and manager. Everything past those five facts is a placeholder. The reading list says list for role. The buddy line says suggested person. The first month's goals are three identical brackets. The file also carries a tip I kept. Don't overload day one. Deep work starts in week two.
Each template blank maps to one kind of team document
Source: how-we-work-index.md (“Start with the first three”), plus team-charter.md
Source: q3-project-list.csv (shipped, owner “unassigned”, “needs an owner”) excerpt
Source: org-chart.csv excerpt
“Fill every blank in this onboarding plan from these four files, and after each line name the file it came from. If no file supports a line, leave the blank and say so.”
The reading list comes from the team's index of how-we-do-things docs, and Tallowmere's index says to start with the first three. What the new hire owns by day thirty comes from last quarter's project list. Look for the small job that shipped and still has no owner. At Tallowmere, that's a returns dashboard. Tom Adeyemi built the returns dashboard, then moved to the pricing team. Who to meet comes from the org chart and from the charter's list of partner teams. So the instruction to Claude is one sentence. Fill every blank in this onboarding plan from these four files, and after each line name the file it came from. If no file supports a line, leave the blank and say so.
A tagged line you can check, and a “my suggestion” line is yours to decide
“owns three of the first-read docs”
“Start with the first three.”
On my run, Claude's draft came back with forty-four lines, and forty of those lines ended with a file name. I checked all forty file names against the files. Thirty-nine held up. One had a wrong detail. The buddy line said Rafael Ortiz owns three of the first-read docs. The index lists three docs to start with, and Rafael owns one of those three. The file name let me find that mistake in seconds. Sorted by where each line came from, the page holds three kinds of line. Twenty-six lines are blanks filled straight from a file. Ten lines are Claude's own picks, and Claude marks them as its suggestions while still naming the file behind each one. Those ten are the buddy, the first task, and the goals for day thirty, sixty and ninety. The last eight lines are missing a source for all or part of what they say, and the doc says so on every one of them.
Four team files and one pasted template go into a single request
| Start with | /docs in the message box |
|---|---|
| Attached | 4 files · charter · project list · how-we-work index · org chart |
| Pasted | Anthropic’s template · one instruction |
| Plans | in beta on Pro, Max, Team, and Enterprise plans |
|---|---|
| Default | on by default on Pro, Max, and Team plans |
| Enterprise | off by default until an owner turns it on |
Claude Docs is Anthropic's new document editor inside Claude. You ask, and Claude writes a doc you can edit and comment on. Anthropic says a doc can draw on your files, your projects and the apps you've connected. I typed slash docs in the message box, attached the four files, and pasted Anthropic's template under the one instruction. The four files are the team charter, last quarter's project list, the how-we-work index and the org chart. Claude Docs is in beta. It's on by default on the Pro, Max and Team plans. On Enterprise, an owner has to switch it on first.
The draft came back with Rafael as buddy and a question about Bea
“Day 1 is a Monday, and the team’s Monday metrics review runs at 10:00 (team-charter.md). That’s the same slot as IT setup.”
The draft took about two minutes, and that time is measured. Claude wrote the word source and a file name after every line it could fill. Claude also left a comment in the margin of the buddy line. The comment said, I suggested Rafael as buddy because he's in Chicago with Dana. Would you rather have Bea, who covers surveys and returns? Claude also caught a clash I hadn't planted. Dana starts on a Monday, and the team's Monday metrics review runs at ten. The template puts IT setup at ten as well.
The first person they lean on and their first owned job are the manager’s to pick
From a file
You decide
Claude’s pick · the manager confirms
Here is where I'd stop and read it myself. Two lines in this plan belong to the manager. The first is the buddy, the colleague a new hire asks the small questions. The org chart lists everyone on the team, and the org chart can't tell you who is generous with their time. So I answered Claude's comment as the manager, with one name. I picked Bea, because Bea wrote the returns taxonomy and Dana's first month is mostly returns work. The returns taxonomy is the team's list of reasons customers send things back. Twenty-six seconds after my reply, Claude answered done, with Bea swapped in at three places in the plan. The second line is what Dana owns by day thirty. Claude picked the returns dashboard that shipped with no owner, and that's the right kind of job. The project list shows what's open. What's safe for a new hire to get wrong in month one is a call for Ines Carvalho.
Export the first draft before you edit, since Claude Docs has no version history
“Version history isn’t available yet.”
Once the plan is right, it leaves Claude Docs. On my account, the Export panel downloads Word, PDF or Markdown, and sends a copy to Google Docs or Notion. Anthropic's help page says version history isn't available yet. So a saved copy of Claude's draft is your only way back to it.
The plan closes on four sentences in the manager’s own words
Skip Claude for the welcome note. The welcome note is four sentences from the manager, and it's the first thing the new hire reads. Every other line in this plan points at a file. The welcome note has no file behind it, because the source is you. I wrote this one by hand in about two minutes. It tells Dana her first Monday opens with the metrics review, and it tells her to ask Bea anything, twice if she needs to. Right above the note, Claude had already listed seven blanks the four files leave empty.
Most of the Claude Docs time goes to checking forty source tags
Both of these times are estimates for this one plan. Built by hand from the four files, the plan takes about forty-five minutes. In Claude Docs the plan takes about fifteen minutes. Only the draft was timed, at about two minutes. About nine of the fifteen minutes go to reading each file name against its file, and that's where the Rafael mistake turned up.
On Monday Dana opens a plan whose lines name a file or mark the gap
Dana Whitcombe starts on the twenty-eighth. Her plan says what to read first and who wrote each doc. The buddy and the first job are the manager's calls, and the welcome note is in the manager's own words. What's the one file on your team you'd want a new hire to read first?
