Urgent.News

What's breaking now, across thousands of outlets.

Tech

How a Security Bug Gave Birth to a Method, Then the Method Caught What the Audit Missed

At 5:27pm on 2026-07-03, I revoked self-write access on 8 sensitive columns of our members table. At 9:42pm the same day, I reverted it — an admin save feature started throwing 403 errors. Locked, then unlocked, same day. That incident became the seed for a method we call the Confrontation-Point Law. I won't claim it's proven — there's exactly one real case so far, and we've deliberately labeled…

On July 3, 2026, the author revoked self-write access on eight sensitive columns of their members table. The same evening, they reverted the lockdown due to a 403 error thrown by an admin save feature. This incident became the foundation for a method called the Confrontation-Point Law. Although untested, this single case is one of the reporter's favorite technical stories, as it caught an oversight that a thorough human audit had missed.

The 07-03 lockdown commit appeared thorough, with the header stating that all five write paths to the users table had been verified. However, the function saveEditMember simultaneously performed a direct write and an RPC call, causing the 403 errors. The auditor read the code, saw the RPC call, and labeled it as legitimate, missing the direct write channel.

The root cause of the issue was not the large number of write points but the auditor's reliance on narrative rather than mechanical verification. The author coined the first rule of the Confrontation-Point Law: every frozen judgment must name its confrontation point, where intent and actual state meet, to prevent harm. Only a machine, like a query to check column privileges, should sit in this seat, not a human relying on narrative.

On July 4, a mechanical re-audit caught three additional write points that the human review had missed, including renewTier, dismissBinding, and importMembers. All eight sensitive columns were re-verdicted, and the three admin write paths were migrated into new SECURITY DEFINER RPCs. The author emphasizes that this method has only one real case so far, labeling it "armed" rather than "battle-tested."

The author also mentions a humorous footnote: during a prompt involving Fable, the AI refused to hallucinate a plausible answer, demonstrating the principle of refusing to answer as a legitimate output. Lastly, the author notes that an unresolved issue exists in Addresses.vue, where a column called addresses does not exist in the database.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

More from Tuesday 8 September →