A security advisory published on GitHub on 19 November 2026 describes a code injection flaw in jane-php/open-api-3, a PHP library that generates API client code from OpenAPI documents. Tracked as CVE-2026-63413, the issue carries a highest reported CVSS score of 9.3, which puts it in the critical range.
The flaw means a malicious or tampered OpenAPI document can cause Jane to write attacker-controlled PHP into the client it generates. That code then runs whenever the generated client is created. A fix is available in version 7.12.0. The issue is not listed in CISA's Known Exploited Vulnerabilities catalogue, and we have seen no reports of attacks so far.

What is affected
The advisory covers the Composer package jane-php/open-api-3. The problem was verified on version 7.11.2 running on PHP 8.1.
- Package:
jane-php/open-api-3(Composer) - Vulnerable: versions earlier than 7.12.0 (confirmed on 7.11.2)
- Fixed: 7.12.0 and later
- Also at risk: any PHP client code already generated by a vulnerable version from an untrusted OpenAPI document
What goes wrong
Jane builds PHP code in a structured way, so ordinary text values in the generated code are properly escaped. The weak spot is the documentation comments (PHPDoc) that Jane writes above each generated method. Those comments include the names of HTTP header parameters taken straight from the OpenAPI document, and that text was written out without any escaping.
A PHP doc comment ends with a specific closing marker. If a header name in the OpenAPI document contains that marker, the comment ends early and whatever follows is treated as real PHP code inside the generated class. The advisory shows that this can be used to add a constructor to the generated Client class. A constructor runs automatically every time the class is instantiated, which most applications do on every API call. The result is remote code execution, classed as CWE-94 (code injection).
Why it matters
Many teams generate PHP SDKs from OpenAPI documents supplied by partners, vendors or public sources. Developers tend to trust generated code and rarely read it line by line. A single crafted header name is enough to hide executable code in a file that looks routine.
Once that code runs inside your application, it has the same access as your app: database credentials, API keys, customer data and the server itself. It would also travel through your normal deployment pipeline, so it could reach production without anyone noticing. The main condition is that the OpenAPI document must be one an attacker can influence.
What to do now
- Update
jane-php/open-api-3to 7.12.0 or later with Composer and commit the updated lock file. - Regenerate every client produced with an older version. Updating the library does not change code that has already been generated and committed.
- Review generated files for anything unexpected in the Client and Endpoint classes, such as a constructor you did not expect, or doc comments that seem to end early around header parameter names.
- Check where your OpenAPI documents come from. Prefer documents from sources you control or trust, keep copies in version control, and review changes before generating code.
- Rotate secrets if you find injected code in a generated client that has run anywhere, and treat that environment as potentially compromised.
The advisory's suggested fix is to neutralise the comment-closing marker in all generated PHPDoc text, including parameter names and descriptions. If you maintain your own code generator or templates, it is worth checking that they handle comment text with the same care.
Not sure if this affects you? Talk to us and we will check your sites and servers and plan the update with you.
Sources
- Jane: Instantiation RCE via header parameter name -> raw PHPDoc */ breakout (jane-php/open-api-3): GitHub Advisories, 19 November 2026
