> For the complete documentation index, see [llms.txt](https://docs.visdom.virtuslab.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.visdom.virtuslab.com/visdom-code-review/what-it-catches.md). # What it catches Generated code fails differently. The names are right, the formatting is clean, the comments are plausible, and the mistake is three levels down. These are the patterns a conventional pipeline and a tired reviewer both miss, and each one is a dedicated lens in the deep layer. **Circular tests.** A test that encodes the implementation rather than the specification. It passes because both sides made the same assumption, so the build stays green while the behaviour is wrong. **Hallucinated APIs.** Calls to methods, fields or endpoints that do not exist in your codebase, or that existed in a version you no longer run. **Convention drift.** Code that works but ignores how your organisation does things: the wrong HTTP client, a bypassed service layer, a pattern the team retired two years ago. Individually trivial, collectively an erosion. **Defensive over-engineering.** Factories, wrappers and abstractions nobody asked for, added because the model has seen a lot of enterprise code and reaches for it by default. The same schema that describes those lenses describes yours. A rule your team keeps repeating in review is a rule that can be written down once and enforced from then on. ## What lands on the pull request One structured comment rather than a wall of them, with findings grouped and linked to the exact line, each naming the rule, the risk level and a concrete fix. Suggested fixes arrive as suggestion blocks a reviewer can apply from the pull request. Findings are deduplicated across layers, and when several layers agree on the same issue that agreement is recorded rather than repeated. Results export in a standard format for code-scanning dashboards and audit pipelines. **Every finding carries a confidence level.** High confidence is reported directly. Low confidence is marked as needing verification and never presented as certain. ## Precision over recall The design doctrine is precision. Past roughly fifteen per cent noise, people stop reading the tool, and a review process nobody reads is worse than none because it looks like coverage. So there are hard caps on how many findings a review may raise, confidence thresholds under which a finding is not shown, and deduplication between layers. Missing a minor issue costs less than training a team to ignore the bot. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.visdom.virtuslab.com/visdom-code-review/what-it-catches.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.