Check the claim against its source#
Dependency comments fall into a few kinds, and each has a source that settles it:
| The comment says | What settles it | Decision if the claim holds | Decision if it fails |
|---|---|---|---|
| This version does not exist | The registry's version list or the artifact URL | Change the version | Dismiss, linking the registry entry |
| This bump has breaking changes | The dependency's changelog and release notes for every version crossed | Change your code, or hold the bump | Dismiss, citing the changelog |
| This range admits an incompatible release | The range syntax and the versions it currently resolves to | Narrow the range, or test the upper end | Dismiss |
| The new version drops a runtime you support | The dependency's engines, requires-python or equivalent, against yours | Hold the bump, or change your support policy on purpose | Dismiss |
| The lockfile and the manifest disagree | A fresh install that rewrites the lockfile | Commit the regenerated lockfile | Dismiss |
| This bump causes a problem for consumers | A reproduction, then the same check on the base branch | Change, if the bump introduced it. Defer to an issue with an upstream trigger, if the base branch already has it | Dismiss |
The registry and changelog answer the first two. Your support policy answers the rest: the oldest runtime you test, the version range you publish, and who consumes your package.
A version the bot said was invented#
In November 2026, a Flutter codelabs pull request regenerated the Firebase "get to know Flutter" steps and moved their Android builds from Gradle 9.3.1 to 9.5.0, the Android Gradle Plugin from 9.1.0 to 9.3.1 and the Kotlin plugin from 2.4.0 to 2.4.20. Gemini Code Assist's review said "non-existent, hallucinated versions were specified for Gradle, the Android Gradle Plugin, and Kotlin". Review
A maintainer approved, and the author merged the pull request the same day. The merged diff keeps gradle-9.5.0-all.zip in each step's wrapper properties. Pull request
The claim is a lookup, so I looked. Gradle's version list at services.gradle.org/versions/all records 9.5.0 with a build time of 28 April 2026, and the distribution URL from the diff returns HTTP 200. Maven Central serves kotlin-gradle-plugin-2.4.20.pom. I could not reach Google's Maven repository from my environment, so I did not confirm the Android Gradle Plugin version. At least two of the three versions exist, and the bot gave the same reason for all three.
A comment saying a version does not exist deserves a dismissal once the registry shows it. The useful reply links the registry entry, which settles the question for anyone who sees the same comment on the next bump.
A real problem the bump did not cause#
A pull request in algoan/pubsub, in September 2026, bumped pino from ^9.14.0 to ^10.3.1 and added @types/node ^24.13.6 as a development dependency. CodeRabbit's review said: "@types/node is a dev dependency, so the ^24.13.6 pin does not control consumer type checking." Review Its point was that the package exposes pino's LoggerOptions type, which reaches thread-stream, whose declarations a consumer on Node 26 types would fail to compile.
The author agreed with the fact and disputed the cause: "Correct that the pin only affects this repo's own build, but this isn't introduced by this PR and" "there's no compatible upstream version to move to". The reply explained that thread-stream@3.2.0, resolved by pino 9 on the base branch, already referenced workerThreads.TransferListItem at line 89, and that removing the exposure "is a breaking public API change and out of scope for a dependency bump." It ended: "Leaving as-is until upstream ships @types/node@26-compatible types." Author's reply The pull request merged on 23 September 2026.
Every part of that reply is checkable. From the npm registry tarballs: thread-stream@3.2.0 references TransferListItem on line 89 of index.d.ts, thread-stream@4.2.0 on line 96, and @types/node@26.0.0 declares no TransferListItem, while @types/node@25.9.3 does. So the bot was right about consumers, and the author was right that the base branch had the same problem.
The deferral also had a trigger, and it has since fired. thread-stream@4.2.1, published on 2 November 2026, no longer mentions TransferListItem. I compiled a one-line import with TypeScript 5.9.3, @types/node@26.0.0 and skipLibCheck: false: against 4.2.0 it fails with TS2694: Namespace '"worker_threads"' has no exported member 'TransferListItem', and against 4.2.1 it compiles. Because pino 10.3.1 and the current 10.4.0 both depend on thread-stream ^4.0.0, consumers who resolve dependencies fresh get the fixed version without a new release of the package. The pull request's lockfile names 4.2.0, so the repository's own build keeps the old declarations until its lockfile is regenerated.
Use the support policy to break ties#
The support policy is the other half of a dependency decision. The algoan diff shows the package's engines field as ^20.19.0 || ^22.13.0 || >=24, and the new thread-stream@4.2.0 entry in the lockfile declares "node": ">=20", so on this axis the bump stays inside the declared range. Had the dependency required Node 22, the bump would have dropped Node 20 users, and that would have been the comment worth acting on.
Write the reply in the same form each time: the claim, the source you checked, the version you checked it at, and the decision. If you defer, name the upstream release that ends the deferral, as the algoan author did. The guide to dismissed findings scores a deferral like this one as valid and deferred, which matters if you are measuring the reviewer. If the bot's comment is about a deprecation in the new version, the deprecation guide covers that check.