Employee Offboarding Checklist to Revoke Access on Day One
Someone on your team hands in their notice, and three weeks later you discover their SSH key still opens the production box.

Someone on your team hands in their notice, and three weeks later you discover their SSH key still opens the production box. That gap is common in teams of five to thirty people, and an employee offboarding checklist is the cheapest way to close it. This guide walks through the order I use as a security writer when a developer, contractor or support person leaves, with the exact accounts, keys and secrets to touch and the timing that keeps work moving.
The goal is simple: by the end of the person's last working day, nothing they personally hold should still grant access to company systems. Not their laptop, not their phone, not a token in a CI variable they created eighteen months ago.
Why an employee offboarding checklist beats memory
Small teams rarely have a central identity system that controls everything. Access tends to grow sideways. A developer gets a GitHub seat, then a cloud console login, then a shared password for the DNS registrar, then a personal API token for the email provider because it was faster than asking. Each of those was a reasonable decision at the time. Together they form a list nobody has ever written down.
When a person leaves, the manager remembers the obvious items: email, chat, the code host. The forgotten ones are the dangerous ones, because they are exactly the accounts that nobody looks at. A written checklist turns the cleanup from a recall exercise into a reading exercise, and it gets better each time someone leaves and you add the item you missed.
Build the access inventory before anyone resigns
The best time to write the list is when nobody is leaving. Spend one hour with a spreadsheet and fill in four columns: system, who has access, how they log in, and who can revoke it. A typical twelve person team ends up with 25 to 40 rows. Expect these categories:
- Identity and communication: the email domain admin, chat workspace, video calls, shared calendars.
- Code and delivery: code host, CI runners, container registry, package registry tokens.
- Infrastructure: cloud consoles, VPS provider, DNS registrar, server SSH users, VPN peers.
- Business tools: billing dashboards, analytics, support desk, the password manager itself.
- Physical items: laptops, hardware security keys, office keys or badges.
Mark every row where access is shared rather than personal. Shared logins are the rows that will cost you time later, because revoking them means changing a credential everyone else also uses.
The order that matters on the last day
Sequence matters more than speed. If you rotate a shared secret before you remove the person from the password manager, they can simply read the new value. Here is the order that avoids that trap.
- Single sign on and email first. Suspend the account rather than deleting it. Suspension blocks logins while keeping mail and files for handover. Most providers also kill active sessions when you suspend, but check that sessions on phones are actually signed out.
- Password manager next. Remove them from every shared vault before touching any shared credential. Note which vault items they could read; that list becomes your rotation queue.
- Code host and CI. Remove the org membership, then search for personal access tokens and deploy keys they created. On many code hosts, removing a member does not delete deploy keys attached to repositories.
- Servers. Delete their line from each authorized_keys file, remove their Unix user, and drop their VPN peer. Our guide to SSH key authentication on a small Linux server shows where those keys usually live.
- Rotate what they could read. Work through the queue from step 2: API keys, SMTP credentials, database passwords, webhook secrets.
- Collect hardware. Laptop, security keys, badges. Then remove the security key from each account it was registered on.
On a prepared team this takes 60 to 90 minutes. Without the inventory it usually takes most of a week, spread across the moments someone notices a leftover login.
Tokens hide in places you forget
The leftover that bites most often is a long lived token stored as a CI variable or inside a cron job on a server. It belongs to the departing person's account, so it keeps working only as long as their account exists. Delete the account and a nightly backup job fails silently. Leave the account and you have a live credential tied to someone outside the company. The fix is to recreate those tokens under a service account before the last day. Our piece on why rotating API keys trips up small teams covers how to swap a key without downtime.
The advice I disagree with: delete everything immediately
A lot of security checklists say to delete the account the moment the person walks out. I think that is wrong for small teams, and the reason is practical. Deleting a mailbox often deletes the files they owned, the calendar invites they organized, and the documents customers were sent links to. I have seen a team lose a year of supplier contracts that way.
Suspend instead, transfer ownership of files and mail to the manager, and set a reminder to delete after 30 days. A suspended account cannot log in, so the security risk is the same. What you gain is time to discover which automations quietly depended on that user. Delete only after the 30 days pass with no surprises.
The other part of that advice I would push back on is the idea that you need to rotate every shared password the instant someone leaves on good terms. Rotate the high value ones the same day: cloud root, registrar, email admin, production database. The low value ones, like a design tool login, can wait for the weekly batch. Spreading the work keeps you from breaking ten things at once at 6 p.m. on a Friday.
Contractors and partial access
Contractors are where the checklist matters most, because their exit often has no formal last day. A project ends, invoices stop, and their access lives on for months. Two habits fix this. First, give every contractor account an expiry date when you create it, even if you have to extend it later. Most code hosts and cloud providers support this, and a calendar reminder works for the rest. Second, never put a contractor in a shared vault; share individual items so the rotation queue stays short.
If your team uses physical security keys, register a contractor's key as a named device so you can find and remove it later. Our guide to hardware security keys for a small team explains how to label keys so this step takes seconds.
Check your work a week later
Seven days after the last day, run a short audit. Search your cloud provider's audit log for the departed user's name and any key fingerprints they used. Grep your servers for their username in /etc/passwd and in every authorized_keys file. Look at the sign in log for your email domain and confirm there are no successful logins after the suspension. If you find something, add it to the inventory so the next exit is cleaner.
Frequently asked questions
How soon should access be removed when an employee leaves?
Remove interactive access, meaning email, single sign on, VPN and SSH, by the end of the last working day. For a sudden or hostile exit, do it before the conversation ends. Shared secrets can then be rotated over the following one to three days in order of value.
Should I delete or suspend a departing employee's account?
Suspend it first. Suspension blocks logins but keeps mail, files and ownership intact so you can hand them over. Delete after about 30 days once you are sure no automation or shared document depends on the account.
What do small teams most often forget during offboarding?
Personal access tokens, deploy keys and CI variables created under the person's account are the usual misses. Shared logins outside the password manager, such as a DNS registrar password in a chat message, come a close second. A written access inventory catches both.
More in Security
Security
DDoS Protection for a Small Server Without a Big Budget
The first time one of my game servers got knocked offline, I spent two hours convinced the host had a network fault.
Security
Locking Down Your Domain Registrar Account the Right Way
A founder wrote to me last spring because his company email had stopped arriving. Not bouncing, just silent.
Security
Leaked Secret in Git and How to Clean It Up the Right Way
You pushed a commit, looked at the diff a minute later, and saw an API key sitting in plain text in a config file.