A liveness audit reported five of my API tokens dead, including the one the live deploy path depends on. Four of the five were completely fine.
I ran a liveness audit across my API tokens and it came back saying five were dead, including the one the live deploy path depends on. Four of those five were completely fine. The audit was calling the wrong verification endpoint: one that only validates user-owned tokens, and which answers "Invalid API Token" for a perfectly healthy account-owned one.
The first time that endpoint lied to me it cost about five minutes. The second time it cost far more, because by then it was wired into an audit and produced a confidently wrong inventory of my own credentials. A checker that reports everything as dead is not describing reality, it is testing wrongly, and that is much harder to notice than a checker that simply errors.
Secretly fixes the half of that problem I actually control: having one canonical local place to resolve a credential. The naming is fixed as product, environment, area and key, so the same value has the same address in every project and no tool has to guess. Values sit in the operating system secret store, while a separate catalogue holds the non-secret facts, such as who owns it, whether setup is done, and when it was last rotated. I can answer whether something exists and whether it is stale without ever reading the value.
It hides values by default. Listing, status and audit print nothing sensitive, revealing one takes an explicit confirmation, and writing an env file refuses a path inside a git repo unless you insist. A wrong passcode fails closed on the authentication tag rather than handing back plausible garbage. None of this is clever. It is just the set of defaults that stops a credential ending up in a screenshot, and defaults are the only control that still works when you are tired.