← All posts
GradingAudit Process

How We Grade a Smart Contract Audit, and Why One Open Critical Means an F

Security Wiz team3 min read

ShareTelegramX

A long audit report is hard for most readers to use. A grade isn't. Think of the health inspection grade in a restaurant window: you don't need to read the inspector's notes to know where things stand. Every audit we do ends in a letter grade from A to F, and the thinking behind it fits on a page.

What goes into it

The grade is built from seven categories:

  1. Code security. Flaws in the contract code itself: risky math, unsafe calls to other contracts, and patterns known to let an attacker withdraw twice.
  2. Access and keys. Who holds the keys, and what happens if one is lost or stolen.
  3. Upgrades and admin power. Can an administrator change the code instantly, create new tokens, or move users' funds?
  4. Outside dependencies. Price feeds, connections to other blockchains, and other apps the product relies on, and what happens when they fail.
  5. Testing. How much of the code is tested, and whether there are automated tests that try thousands of random actions to break its core rules.
  6. Website and operations. The ordinary web software and systems around the contracts.
  7. Documentation and openness. A written description of how it should work, published code anyone can check, and admin powers that are disclosed.

Each category starts at full marks. It loses points for each problem we find, more for serious ones, and for good practices that are missing. For example, if an administrator can drain funds and there is no waiting period before their changes take effect, the admin category loses points even though no single line of code is wrong.

Severity comes from a public scale

Problems in the contract code are scored with OVSS, our open scoring system. It rates each finding from a short list of inputs, and anyone can use those inputs to recalculate the score. Problems in ordinary web software use CVSS 4.0, the standard scoring system for traditional software. The grade is built from those scores, so it traces back to specific findings and not to a gut feeling.

The rules that stop one bug hiding

An average can hide a disaster. A project could score well on testing, documentation and dependencies and still ship a flaw that empties the vault. So the grade has hard limits:

  • Any Critical finding that is still open or acknowledged means the grade is F.
  • Any High finding that is still open or acknowledged means the grade can't be higher than C.

"Acknowledged" means you chose to accept the risk without fixing it. It counts the same as open, because the risk is the same. A finding we confirm as fixed on retest no longer counts against you. The point of the grade is to reflect the code as it will actually go live.

A grade has a shelf life

The grade applies to one exact version of the code and one launch of it. If the live code changes, or who controls it changes, the grade is out of date until we review it again. That's what makes it useful to your users. It says something specific that can be checked, where the word "audited" on its own carries no date and no version.

Why a grade at all

  • Readers who can't read code (investors, partners, your community) still get a clear signal.
  • Teams get a target. "What's blocking an A?" has a concrete answer: fix the open findings in the lowest-scoring categories.
  • It rewards fixing over arguing. Because a confirmed fix removes the penalty, the incentive is to fix and retest.

We haven't published the exact weights and deductions, because they're still being tuned against real audits. What is settled is the structure: categories, hard limits, and a scale you can check.

To see how your project would score, request an audit, or read how our process works.

ShareTelegramX

Shipping something onchain?

Send us your repo and we’ll return a quote within 24 hours.

Request an Audit