The Framework Count That Cannot Drift
TheAuditor recognizes 138 named frameworks, and it is not a marketing number. A CI test pins the live registry to an explicit frozen set, so the count cannot change unless someone changes the frozen list in the same commit.
Every scanner advertises a framework count, and every framework count is a number in a README that nobody re-checks after it is written. Frameworks get added, renamed, and quietly dropped, and the marketing figure drifts away from what the tool actually does. Six months later the box says one thing and the engine does another, and you have no way to know.
Ours is 138, and the reason we will put that number in writing is that the build will not let it drift.
A number the build enforces
TheAuditor recognizes 138 named frameworks. A test in continuous integration pins the live registry against an explicit, 138-member frozen set and asserts they match exactly. Add a framework and the build breaks. Remove one and the build breaks. The only way to change the count is to change the frozen list in the same commit, in front of a reviewer, on purpose. A pre-commit check and a CI lint stand in front of that test to catch drift before it ever reaches the assertion.
The effect is simple: the number cannot move by accident. It is not a claim we maintain by remembering to update a doc. It is a claim the pipeline maintains for us, and refuses to let us break.
Recognized, and fully declared
We draw one honest line inside that 138. Every one of them is recognized. Of the 138, 76 also ship a full behavioral declaration, the deeper description that lets the engine reason about how a framework routes requests, handles input, and talks to data. The remaining ones are recognized today and get their fuller treatment as the work lands.
We report that split rather than blurring it, because “recognized” and “fully modeled” are different promises, and you deserve to know which one a given framework has. A count that hides a second tier underneath it is exactly the kind of number this whole approach exists to prevent.
Why a number you can hold us to matters
Coverage claims are usually unfalsifiable, which is another way of saying useless. A number that the build itself keeps honest is one you can actually hold us to: it means what it says on the day you read it, and it cannot quietly rot between releases. That is the whole point of pinning it. Trust is not us asking you to believe the figure. It is the figure being unable to lie.
Honest scope note
Recognizing 138 frameworks is a statement about breadth, not a guarantee that every framework is modeled to the same depth, which is exactly why we publish the recognized-versus-declared split instead of one flat number. Where a framework is recognized but not yet fully declared, the engine says so rather than pretending to a depth it does not have. What is proven is measured against a yardstick at BenchProctor, whose answer key the tool is never tuned to.
Where this sits
A coverage number the build keeps honest is the same discipline that runs through Code Reality Labs: claims you can check, not claims you take on faith. TheAuditor is in final commercial release preparation and ships when its hardening checks pass. Subscribe on the main site for launch news.
Was this useful?