All posts

June 16, 2026 · 4 min read

Q&A: Do I Need a Security Check Before Every Feature, or Just Before Launch?

A version of this question comes up often from people building quickly, adding features week over week: do I need to check security on every single one of these, or is it enough to do one real pass right before I open things up to actual users?

The honest answer: it depends what the feature touches

Not every feature carries the same risk. A change to how a button is styled, or how a page is laid out, doesn't need its own security review. A feature that introduces a new way to access, upload, share, or expose data does, because that's exactly where the categories of issue I write about most often tend to show up: a new sharing feature that skips an ownership check, a new upload path with no validation, a new API endpoint that doesn't verify who's calling it.

A useful rule of thumb

If a feature touches user data, changes what's shared with whom, adds a new way to upload or fetch something, or adds a new API route, it's worth a quick, targeted look specifically at that feature before it ships, not a full audit, just checking the specific new surface area it introduces. If it doesn't touch any of that, one broader pre-launch check is genuinely enough.

Why waiting until launch alone is risky

The risk with checking only once, right before launch, is that by then a real number of features have already accumulated, each one a potential place something was missed, and a single pass at the end has to catch all of them at once. Checking the higher-risk features as they're built spreads that work out and catches issues while the context of what changed is still fresh, rather than all at once against a much larger surface.

What this looks like in practice

  • Low-risk changes (styling, copy, layout): no dedicated check needed.
  • Medium-risk changes (a new form, a new page): worth a quick manual look at that specific feature.
  • High-risk changes (sharing, uploads, new API routes, anything touching another user's data): worth a targeted check before that feature ships, not after.
  • Before opening up to real users regardless: one broader pass across the whole app, even if individual features were already checked along the way.

The short version: not every feature, but the ones that change who can see or touch what, yes, checked as you go rather than saved up for one big pass at the end.

Related reading

Harbova is a security service for apps built with AI tools. Start with a free scan, and if it finds something serious, we can fix it and prove it is closed.