Security Consult with Michael (@hackrowd) header
Security Consult with Michael (@hackrowd) profile picture

Senior Security Engineer | Cloud Security (Azure) · DevSecOps · ISO 27001 · ISO 22301 · ISO 20000 · PCI DSS · SWIFT CSP | Securing Cross-Border Payments

Nigeria
Joined December 2019
Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)Photo by Security Consult with Michael (@hackrowd)
Congrats on the launch. The question your buyers' security teams will ask first: can text in a Slack thread or Jira ticket steer the agent? Scope each run's token to one repo and branch, and treat ticket content as untrusted input, not instructions.
16
Replying to @dikesa_dev
If the agent can read, write, delete and send, the guardian becomes the most valuable target in the stack. Worth publishing how it fails: when the policy engine is down or slow, do tool calls fail open or fail closed? SOC 2 buyers will ask that first.
4
New week. Pick one item from your security list and finish it before Friday.
18
OWASP API in plain words: Unrestricted access to sensitive business flows. Bots buy all the stock, farm the referral bonus or book every slot. Add limits that match real human behaviour.
19
Me: writes a clear, polite bug report. The ticket, six weeks later: "Status: Needs triage."
29
Opinion: if a security control annoys users every day, they will find a way around it. Design controls people can live with.
27
Dev habit: develop with a low-privilege test user, not an admin. You'll spot missing permission checks much sooner.
15
Careers: read a company's security pages and engineering blog before the interview. Knowing their stack makes your answers specific.
19
Question: do you use passkeys anywhere yet? Which app made it easy?
22
Replying to @techaurchai
Caching and background jobs. They run outside the request, so the 'current tenant' quietly isn't there. Pass the tenant ID explicitly into every job and cache key, and add a test where tenant A's job runs while tenant B is the one logged in.
1
1
7
Security awareness poster: "Think before you click." The poster: on the wall right beside the sticky note with the shared password.
25
Replying to @LDT0545
Solid list. One to add under #1: use a second account on a different team or org, not just a second user. A lot of real leaks are cross-tenant: the ID check passes because both users exist, but nothing checks they belong to the same tenant.
13
Replying to @GaneshKompella
What works on small teams: whoever merges owns it, with one rule the AI can't skip. Every new table ships with its RLS policy plus a test that a second user gets nothing back. No test in the PR, no merge.
1
1
11
Myth: "Long passwords are hard to remember." Fact: four random words make a passphrase that is long, strong and easier to recall than P@ssw0rd!.
21
Code review: check that every new route sits behind the auth middleware. One route registered outside it is an open door.
23
Founder checklist: you can list every admin of your cloud, domain, email and payments accounts from memory, and the list is short.
17
Cloud: turn on object lock or delete protection for your backups. Ransomware crews go after backups first.
33
Walkthrough: signup says "This email is already registered". Try a list of emails and you learn who uses the app. Fix: same response for everyone, and email the owner instead.
20
"Can you make it secure but also not change anything?" Security engineers: a long pause.
21
AI apps: don't let the model decide who the user is. Identity comes from your auth system and is passed to tools, never inferred from the chat.
22