oapi-codegen: OpenAPI Server Description Escapes Generated Go Comment and Injects Executable Code

Summary

## Summary

The vulnerability in oapi-codegen seems to be similar with CVE-2026-22785, which is a generated-code injection issue where untrusted OpenAPI `summary` text is embedded into generated TypeScript MCP server source without proper escaping. `oapi-codegen` has a similar vulnerability in its server URL generator: untrusted OpenAPI `servers[].description` text is inserted into a generated Go line comment without normalizing embedded newlines. A crafted description can break out of the comment, add imports through `goimports`, and emit executable Go declarations into the generated package.

> [!NOTE] > A vulnerability like this requires that it is missed in code review **and** that you then call the malicious method. > > Using an `init()` function could be enough to not require a direct call to the code, and instead rely on you importing the package, but either way, code review should be performed before any `oapi-codegen` generated code is executed. > > We **strongly recommend** all users to be reviewing changes to their generated code before they execute anything within it, to protect against supply chain attacks or malicious injected code. > > This is also why we recommend `oapi-codegen` generated code is committed to source control.

## Details The vulnerable sink is in `pkg/codegen/templates/server-urls.tmpl`.

```gotemplate // {{ .GoName }} defines the Server URL for {{ if len .OAPISchema.Description }}{{ .OAPISchema.Description }}{{ else }}{{ .OAPISchema.URL }}{{ end }} const {{ .GoName}} = "{{ .OAPISchema.URL }}" ```

This template assumes the OpenAPI server description remains inside a single Go line comment. However, OpenAPI descriptions are attacker-controlled strings and may contain newlines. Once a newline is present, the next line is no longer part of the comment.

The same raw description is also used in the function form of server URL generation:

```gotemplate // New{{ .GoName }} constructs the Server URL for {{ .OAPISchema.Description }}, with the provided variables. func New{{ .GoName }}({{ .NewServerFunctionParams }}) (string, error) { ```

Identifier generation does not protect this sink. In `pkg/codegen/server_urls.go`, the description is normalized only for the generated Go identifier:

```go suffix := server.Description if suffix == "" { suffix = nameNormalizer(server.URL) } name = serverURLPrefix + UppercaseFirstCharacter(suffix) name = nameNormalizer(name) ```

The identifier is sanitized, but the raw `server.Description` is still rendered in the comment template. This leaves the code-generation context vulnerable.

The generated file is then formatted with `goimports` in `pkg/codegen/codegen.go`:

```go goCode := SanitizeCode(buf.String())

outBytes, err := imports.Process(opts.PackageName+".go", []byte(goCode), nil) ```

`SanitizeCode` only removes byte-order marks:

```go func SanitizeCode(goCode string) string { return strings.ReplaceAll(goCode, "\uFEFF", "") } ```

It does not escape comments, replace newlines, or otherwise serialize untrusted text for a Go source-code context. As a result, attacker-controlled source can be preserved and formatted as valid Go.

### How to Reproduce The attacker-controlled input is an OpenAPI document whose `servers[].description` contains a newline followed by Go declarations:

```yaml openapi: "3.0.0" info: title: oapi-codegen server URL description injection version: "1.0.0" servers: - url: https://api.example.com description: | benign var _ = func() int { panic("oapi-codegen generated-code execution") return 0 }() // paths: {} ```

Generate Go source with server URL generation enabled. No special local path or helper file is required for the vulnerability; the malicious description is copied into the generated source-code context.

The generated source contains attacker-controlled executable code:

```go // ServerUrlBenignvarFuncIntPanicOapiCodegenGeneratedCodeExecutionReturn0 defines the Server URL for benign var _ = func() int { panic("oapi-codegen generated-code execution") return 0 }()

//

const ServerUrlBenignvarFuncIntPanicOapiCodegenGeneratedCodeExecutionReturn0 = "https://api.example.com" ```

## Impact

An attacker who can supply or influence an OpenAPI document consumed by `oapi-codegen` can inject arbitrary Go source into the generated package. In common API-client/server generation workflows, this can lead to build-time or runtime code execution in developer machines, CI systems, or downstream applications that trust generated code.