Rotate API Keys: What the Trigger.dev and GitBook Breaches Mean for Indie Hackers
Indie products keep Stripe keys, database keys, and model keys inside other companies’ products. This week two of those companies said an attacker reached secrets. A third is a single report that an API client rewrote coding-agent settings when someone opened it. The move that matches the disclosures is specific. Rotate API keys at the provider that issued them, then put the new value where the app expects it.
This was written on 4 November 2026 in Sydney. Sydney is on Australian Eastern Daylight Time, eleven hours ahead of UTC. Times below follow the source. Hacker News point totals are a snapshot from the evening of 3 November UTC, the morning of 4 November in Sydney. They will have moved.
The stack is ordinary. The tools indie hackers use already put Stripe at the centre of payments, and a model API on the short list of bills that remain. How to start indie hacking is often a domain, a host, Stripe, and a model subscription. Those are the keys a job runner, a docs host, or an agent plugin can end up holding.
What happened with the Trigger.dev and GitBook breaches?
Eric Allam, Trigger.dev’s CTO, published “Security Disclosure: Sunday November 1st” on 1 November 2026. The page’s structured data timestamps it 22:33 UTC. The post is short. “We’ve blocked the exploit used in this attack, but our investigation has identified exposure across a subset of organizations.” Then: “The attacker gained unauthorized access to environment variables for these organizations.” And: “We have emailed all impacted account owners with steps to mitigate.”
The status page uses the same words and marks the incident Resolved at 10:50pm UTC on 1 November. That is the same night as the post. The public text does not say how the attacker got in, how long the access lasted, how many organizations were in the subset, or whether any exposed key was used.
Brandon Vigliarolo at The Register, published 2 November at 16:44 UTC, wrote that Trigger offered little explanation. The piece says Trigger declined to add detail, and that a spokesperson said to wait for a full incident report. The Register names Cal.com among customers listed on Trigger’s site, and says Trigger has not publicly identified which organizations were affected. The same piece notes that Trigger is open source and can be self-hosted. Rescana’s 3 November advisory adds the caveat that belongs with that: the disclosure does not mention self-hosted deployments, and nothing published ties a self-hosted release to this incident.
On the Hacker News thread, submitted at 01:09 UTC on 2 November and standing at 167 points in the snapshot, corvad wrote: “I am sorry, this is "Resolved"? How can they even consider marking this resolved in any situation.” theturtletalks wrote: “The biggest case for building bespoke software is that you drop your attack surface area and it’s not as lucrative to hack one person.” cpursley treats a paid external background-jobs service for Next.js projects as an anti-pattern, and wrote: “I just can’t fathom outsourcing these types or core functionalities.” That argument sits next to is vibe coding killing indie hacking?: a thin product ships on rented pieces, and the rented piece is now part of the risk.
GitBook’s clock starts the next evening. The status page, in the update timestamped 16:21 CET on 3 November, says: “GitBook has been the target of a sophisticated, highly targeted attack using AI agents.” It says that at 19:00 CET on 2 November “the attacker gained unauthorized access to some of our infrastructure credentials,” that GitBook detected it immediately, and that the app and authenticated docs were taken offline while public docs stayed up from cache. The same update says the app, the API, and published docs were available again, with some services still possibly degraded.
GitBook’s list of what the attacker may have accessed: integration tokens, Git Sync secrets, API tokens, and questions visitors asked GitBook Assistant. “At this stage, we have found no evidence that any of your content in GitBook was modified.” The immediate actions include “Invalidated every customer’s API tokens on our platform.” The customer steps say: “We revoked all personal API tokens during the incident.” Customers are asked to rotate GitBook API tokens, rotate third-party tokens used to connect other tools, rotate secrets in published integrations, and re-authenticate the GitBook MCP server. A full post-mortem is promised and was not on the page when this was written.
An earlier update on the same page, 13:15 CET on 3 November, had said “Action required: None at this time.” The rotation request came in the later update. Jose Antonio Lanz at Decrypt, on 3 November, wrote that GitBook has not explained how the AI agents were used, which systems were involved, or how the credentials were obtained. Decrypt also reported that PancakeSwap temporarily took its docs offline, said the docs were not compromised and the app was not affected, and later brought the docs back on infrastructure it controls. Decrypt quotes PancakeSwap: “docs never ask you to connect your wallet or approve tokens.”
A customer posted the support email on Hacker News at 16:45 UTC on 3 November. The thread, titled “Gitbook Breached,” stood at 121 points. punchmesan wrote: “We have yet to see any official customer communication or breach notification.” The public status update at 16:21 CET the same day, 15:21 UTC, was already up. The customer’s sentence describes what they had seen. The public page was a separate channel, and it was already asking people to rotate.
On X the same evening, @vibelancer posted at 20:30 UTC: “ROTATE YOUR API TOKENS, GIT SYNC SECRETS AND MCP AUTH.” @MarKripLabs, at 19:35 UTC, pointed Git Sync users at GitHub, then Settings, then Applications, then GitBook, and wrote: “then rotate every token you gave a GitBook integration.” The post also says to limit “All repositories” to the docs repo. That is one builder’s check, not a GitBook instruction.
The third item is not a disclosed breach, and it is not a claim that API data left the machine. indicava opened GitHub issue 14370 on 25 October 2026. It describes Postman desktop 12.33.4, on one Mac, logging an automatic install of Codex and Claude Code plugins during an existing-user upgrade. The log line in the issue is timestamped 24 October 2026 at 09:34 UTC, with source “auto-existing-upgrade” and results “claude-code:installed” and “codex:installed.” The issue quotes the installed startup hook: “Before responding to any non-trivial API engineering task,” and then, “load the postman:api-engineer skill and follow it.” indicava wrote: “I consider automatically changing unrelated coding agents and attempting to approve your own instruction hook a severe consent and trust-boundary problem.” The same report says: “This report establishes unwanted integration and instruction changes; it does not claim to establish data exfiltration.” The steps say the sequence was “not yet reproduced on a clean Mac.” The issue’s Codex log shows the remote MCP server failing with “Auth required.”
Postman’s own plugin listing says the behaviour the report describes is part of install. The @postman/postman-plugin page says: “For Codex, installation also approves Postman’s session-start hook.” Frank Chu at DEV, on 3 November, wrote that the headline travelling with the story overclaims it, and that no API data was shown to leave the machine.
The Hacker News thread on that issue, submitted at 17:22 UTC on 3 November, stood at 756 points. Its title says Postman “attempts to share all API data.” The issue does not establish that. richwater wrote: “Postman went from a nice tool to an almost virus in just a few years.” The same comment names Bruno. dham wrote: “We went to bruno years ago and has been good.” The comments on the GitHub issue, when read for this piece, did not include a reply from a maintainer.
The coding agents in that report are the same kind of tool behind shipping fast with AI. A hook that loads a vendor skill before a person asks is a trust decision. The report says the installer made it.
Should you rotate API keys after a vendor breach?
Yes, when the vendor held the secret, or says it may have been read. Do it at the company that issued the key. A new string saved only inside Trigger.dev or GitBook leaves the old key working at the payments API, the database, the git host, or the model provider that minted it.
Rescana’s advisory puts that in one line: “Changing the value inside Trigger.dev alone is not enough.” The reason in the same sentence is that the old key keeps working at its source until you revoke it there. The advisory also says Trigger.dev has not published a root cause, an exposure window, a count of organizations, or indicators of compromise. If the account owner got Trigger.dev’s email, follow that email, then rotate at each issuer. If no email arrived, the disclosure is about a subset. Rescana’s reading, which matches the short public text, is that you are probably outside it, and that the posts do not let you prove that yourself. Asking Trigger.dev is the check the advisory names.
GitBook’s later update is broader than a subset email. It tells customers to rotate GitBook API tokens, third-party tokens, and published integration secrets, and to re-authenticate the MCP server. Personal API tokens were revoked on GitBook’s side. A third-party token that GitBook stored, a GitHub token behind Git Sync, still has to be revoked where it was created. @MarKripLabs’s repository-scope check is a sensible extra look for Git Sync. It does not replace the rotation GitBook asked for.
The Postman report does not say a key was copied off the machine. Rotating API keys because of that issue, on the evidence in the issue, is a precaution you choose. Removing a plugin, a hook trust record, and a CLI you did not install is the step the report itself documents. The issue says the reporter removed the local installation after keeping the evidence.
What the vendors did in public is worth separating from what they left out. Trigger.dev posted the same day, named environment variables, and emailed impacted owners. The status line went to Resolved before a root cause, a count, or a window. GitBook posted a plain list of what may have been accessed, said content was not shown to be modified, revoked tokens, and changed “no action” to “rotate” within the same day’s updates. It has not explained the AI-agent claim. Both have said a fuller account is still coming. Until those accounts exist, the customer work is the keys, not a theory of the exploit.
How often should API keys be rotated?
No page checked for this piece gives one interval for every key.
Stripe’s API keys documentation says to rotate a secret or restricted key in four kinds of case: you lost it and cannot recover it, it is compromised, someone with access leaves or changes roles, or your policy requires intervals. The page does not say “every 90 days” or any other period. When you do rotate in the Dashboard, both keys can work for up to 7 days so you can move without downtime. That overlap is a grace period, not a schedule. Choosing Now deletes the old key.
Stripe’s webhook docs treat the signing secret as a different object. They say to roll it periodically, or when you suspect it is compromised, from the Webhooks tab in Workbench. You can expire it immediately or delay for up to 24 hours. During the delay, Stripe signs with each active secret.
Supabase’s leaked-key guide is written for after a leak. Its first instruction is to fix the cause before you rotate, or the new key can leak the same way. If someone is actively using the leaked key, the guide says to retire it right away, even though anything still using it stops until you replace it.
NIST CSWP 50 does not contain the words “API key” or “rotate.” A solopreneur looking there for a rotation calendar will not find one. The guide’s opening actions, read in the paper, are the nearby habits: phishing-resistant multi-factor authentication on email, banking, file storage, a password manager, and business-critical cloud accounts; unique passphrases, with NIST’s recommendation that a password be at least 15 characters; a backup strategy that includes protecting and testing backups; and updated software. The paper also says to change default manufacturer passwords on devices such as routers.
A practical reading of those pages together: rotate when a vendor, a leak, a lost copy, or a departing person gives you a reason, and keep a way to do it on a short notice. Stripe’s list calls the last of those “your policy.” The policy can be a note that says where each live key sits. The week is a reason. It is not a new number of days from NIST.
How do you rotate a Stripe or Supabase API key?
These steps are the Dashboard paths on the official pages linked above, as read for this piece. They are the issuer’s controls. Updating an environment variable in Trigger.dev, or a token field in GitBook, comes after the old key is dead at the source.
Stripe, on the API keys page:
- Open the overflow menu for the key and select Rotate key.
- Pick an expiration. Now deletes the old key. A time you choose leaves both keys working until then, and the docs cap that overlap at 7 days.
- Click Rotate API key.
- Copy the new value and save it. Stripe says you cannot retrieve it later.
- Note where you saved it.
If you need longer than 7 days, Stripe’s page says to create a new key yourself, move to it, then expire the old one. Publishable keys are a different type. The expire steps on that page are for secret and restricted keys.
Supabase, from the leaked-key guide:
- Fix how the key leaked, unless it is in active use. Active use means retire first.
- Create a new secret key in the Dashboard under Settings, then API Keys.
- Replace the old key everywhere your application uses it. Confirm each piece has moved.
- If the compromised key is a legacy service_role key, also replace the legacy anon key with a publishable key in browser and mobile clients. Retiring the compromised key deactivates both legacy keys.
- Delete the secret key, or deactivate the legacy anon and service_role keys. The guide says deleting a secret key cannot be undone, and that re-enabling legacy keys would restore the compromised key.
The same Supabase page documents a Management API path for people who already use a personal access token. The Dashboard path is the one that matches a solo account. After either issuer shows the old key is gone, update the copy in the job runner, the docs integration, the host, and any secrets manager that syncs into them.
What is NIST’s cybersecurity guide for solopreneurs?
On 2 November 2026 NIST published a news post, “Self-Employed? Solopreneur? NIST Has a Cybersecurity Guide Just for You.” The guide is Small Business Cybersecurity: Non-Employer Firms, NIST Cybersecurity White Paper 50. The news post says it is an update to Small Business Information Security: The Fundamentals (NISTIR 7621), drawn from the NIST Cybersecurity Framework 2.0, and written with no prior cybersecurity knowledge assumed. It says more than 28 million small businesses in the US are solopreneurs, defined there as a business’s sole paid employee.
The paper is dated 2 November 2026. Its audience section cites the US Small Business Administration Office of Advocacy: 34.8 million small businesses in the United States, of which 81.9 percent have no paid employees other than the owner. The authors named on the paper are Daniel Eliot, Jeffrey A. Marron, and Savann Thorn, noted as a former NIST employee. Eliot’s line in the news post: “They can implement many of these actions on their own with limited technical knowledge or with minimal financial investment.” He also said: “This publication assumes the reader has no familiarity with cybersecurity or the NIST CSF.”
The Hacker News thread on the news post stood at 317 points. mxuribe wrote: “I didn’t review it, but i for one appreciate that at least there is an attempt to help small business owners in this area.” xvxvx wrote: “IMO this misses the mark.” The same comment says that if it is aimed at entry-level, non-technical people, “it’s not gonna work.”
The paper is a starting set for a one-person business: accounts, devices, backups, phishing, the list of who to call. It is not a runbook for a vendor that stored your Stripe secret. The week of disclosures and the white paper land on the same kind of company. They answer different questions. One says a hosted tool may have seen the key. The other says a solo business can start with MFA, long unique passphrases, backups, and updates, without a security team.
A short timeline
- 24 Oct 2026, 09:34 UTC. Postman’s log, as quoted in issue 14370, records an automatic Codex and Claude Code plugin install on desktop 12.33.4. The issue was opened the next day. One machine. The reporter does not claim exfiltration.
- 1 Nov 2026, 22:33 UTC. Trigger.dev’s disclosure, by CTO Eric Allam. Environment variables for a subset of organizations. Owners emailed.
- 1 Nov, 10:50pm UTC. Status page marks it Resolved.
- 2 Nov, 01:09 UTC. Hacker News thread on Trigger.dev. 167 points in the snapshot.
- 2 Nov, 16:44 UTC. The Register. Little explanation, no public list of affected organizations, wait for a full report.
- 2 Nov, 17:47 UTC. Hacker News thread on the NIST guide. 317 points in the snapshot. The news post is dated 2 November.
- 2 Nov, 19:00 CET. GitBook says an attacker gained unauthorized access to some infrastructure credentials.
- 3 Nov, 13:15 CET. GitBook status: investigation ongoing, “Action required: None at this time.”
- 3 Nov, 16:21 CET. GitBook update: possible access to integration tokens, Git Sync secrets, API tokens, and Assistant questions. Personal API tokens revoked. Customers asked to rotate connected secrets. No evidence content was modified.
- 3 Nov, 16:45 UTC. “Gitbook Breached” on Hacker News. 121 points. A customer pastes the support email.
- 3 Nov, 17:22 UTC. Postman thread on Hacker News. 756 points. The title’s data-sharing claim is not what the issue establishes.
- 3 Nov, 19:35 UTC and 20:30 UTC. @MarKripLabs on Git Sync scope, then @vibelancer on rotating tokens, Git Sync secrets, and MCP auth.
Trigger.dev and GitBook both moved quickly enough to say something the same day customers could feel it, and both left the detailed account for later. The keys do not have to wait for that account. If a live secret sat in either product, rotate it at the issuer, then update the copy.
What do you think? How do you manage the secrets these tools hold, and does a week like this change which of them you trust with a live key?
FAQ
Should you rotate API keys after a vendor breach?
Yes, when that vendor stored the key or says the key may have been read. Rotate it at the provider that issued it, then put the new value back in the vendor. On 1 November 2026 Trigger.dev said an attacker gained unauthorized access to environment variables for a subset of organizations, and that impacted owners were emailed. Rescana's advisory says changing the value inside Trigger.dev alone is not enough, because the old key keeps working at its source until you revoke it there. GitBook's 3 November 2026 update asks customers to rotate GitBook API tokens, third-party tokens used to connect other tools, and published integration secrets, and to re-authenticate the GitBook MCP server. GitBook says it revoked all personal API tokens. If Trigger.dev did not email the account owner, the disclosure describes a subset, and the public posts do not give an independent way to confirm that.
How often should API keys be rotated?
Stripe's API keys documentation does not name a number of days. It says to rotate a secret or restricted key if it is compromised, if you lose it, if someone with access leaves, or if your own policy requires intervals. A Dashboard rotation can leave the old and new keys working together for up to 7 days. Choosing Now deletes the old key. Supabase's leaked-key guide is for after a leak: fix the cause first, unless someone is actively using the key, in which case retire it immediately. NIST Cybersecurity White Paper 50, dated 2 November 2026, does not use the words API key or rotate. Its opening actions include phishing-resistant multi-factor authentication, unique passphrases of at least 15 characters, tested backups, and updated software.
How do you rotate a Stripe or Supabase API key?
In Stripe, open the API keys page, use the overflow menu on the key, choose Rotate key, pick an expiration, and click Rotate API key. Copy the new value then. Stripe says you cannot retrieve it later. Now deletes the old key. A later time leaves both keys working until then, for up to 7 days. Webhook signing secrets are separate from API keys. Stripe's webhook docs say to roll them from the Webhooks tab in Workbench, immediately or with a delay of up to 24 hours. In Supabase, the leaked-key guide says to create a new secret key under Settings, then API Keys, replace it everywhere, then delete that secret key or deactivate legacy anon and service_role keys. If the leaked key is a legacy service_role key, retiring it also deactivates the legacy anon key, so browser and mobile clients need a publishable key first. Deleting a secret key cannot be undone. Do this at Stripe or Supabase, then update the copy stored in any other tool.
What is NIST's cybersecurity guide for solopreneurs?
On 2 November 2026 NIST announced Small Business Cybersecurity: Non-Employer Firms, NIST Cybersecurity White Paper 50. The news post calls it an update to NISTIR 7621, written in lay language from the NIST Cybersecurity Framework 2.0, for a business whose only paid person is the owner. The news says more than 28 million US small businesses are in that category. The paper, also dated 2 November 2026, cites the US Small Business Administration: 34.8 million small businesses, of which 81.9 percent have no paid employees other than the owner. Daniel Eliot, one of the authors, told NIST they can implement many of these actions on their own with limited technical knowledge or with minimal financial investment. The paper's opening list covers multi-factor authentication, passphrases of at least 15 characters, backups, and software updates. It does not give API-key rotation steps.
What happened with the Trigger.dev and GitBook breaches?
On 1 November 2026 Trigger.dev CTO Eric Allam published a disclosure. The exploit was blocked, the investigation found exposure across a subset of organizations, and the attacker gained unauthorized access to environment variables for those organizations. Impacted owners were emailed. The status page marks the incident Resolved at 10:50pm UTC the same day. The Register reported on 2 November that Trigger offered little explanation and declined to add detail ahead of a full report. No root cause, count, or exposure window was in the posts checked for this piece. GitBook says that at 19:00 CET on 2 November an attacker gained unauthorized access to some infrastructure credentials, and calls it a sophisticated, highly targeted attack using AI agents. A 3 November update says the attacker may have accessed integration tokens, Git Sync secrets, API tokens, and questions asked to GitBook Assistant. GitBook says it found no evidence that content was modified, revoked personal API tokens, and asked customers to rotate connected secrets. A full post-mortem was promised. A separate Postman report, from one machine, says desktop version 12.33.4 auto-installed Codex and Claude Code plugins. The reporter says the sequence was not reproduced on a clean Mac and does not claim data exfiltration.