CVE-2026-64650 @ai-sdk/harness-codex vulnerability
The `@ai-sdk/harness-opencode` tool is an HarnessV1 adapter backed by @openai/codex-sdk, which drives the codex command line interface. Prior to version 1.0.29, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (the Codex CLI shim). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools, including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.29 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the Codex harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code.
Browse Linux security risksQuick answer
vercel @ai-sdk/harness-codex should be reviewed and updated if it matches the affected versions. The recommended fix is to apply the vendor-supported patched version or the mitigation steps below, then retest the public website with Fixnx.
Who is affected
Affected versions
- < 1.0.29
Fixed versions
- Apply the latest vendor-supported patched version.
How to fix it
CVE-2026-64650 affects @ai-sdk/harness-codex and involves authorization failure. Successful exploitation may let a lower-privileged user view or change data outside their allowed scope. Prioritize deployments that process untrusted input, expose the affected feature, or hold sensitive data. Upgrade @ai-sdk/harness-codex to 1.0.29 or a later vendor-supported fixed release.
- Inventory every deployment, device, package, plugin, container, build, and environment that uses @ai-sdk/harness-codex.
- Compare the installed version with the recorded affected range: < 1.0.29.
- Upgrade @ai-sdk/harness-codex to 1.0.29 or a later vendor-supported fixed release.
- Until patched, restrict the affected API or workflow to trusted administrators and enforce an additional access-control layer at the proxy or network boundary.
- Review configuration, roles, tokens, network paths, and exposed endpoints connected to @ai-sdk/harness-codex; remove access that is not required.
- Review authentication, API, audit, and object-change logs for cross-user access, bypassed checks, or unexpected privileged actions.
- If exploitation is suspected, isolate affected assets, preserve evidence, rotate exposed credentials or keys, restore trusted data, and rebuild compromised systems before returning them to service.
Scan now. Google sign-in is only needed to unlock fix guidance.
Verify the fix
- Confirm every affected asset now runs the fixed vendor-supported release or has the documented mitigation for CVE-2026-64650 in place.
- Use a safe regression test to confirm the affected authorization failure path no longer produces the vulnerable behavior.
- Confirm temporary access controls, endpoint restrictions, network blocks, and least-privilege settings remain effective after the update.
- Review post-remediation logs and rerun Fixnx plus the relevant package, dependency, firmware, browser, or product-specific security check.
- Document affected assets, versions, changes, validation evidence, reviewed logs, and any credential rotation, cleanup, or replacement completed.
Related categories
Related security risks
More published guidance from the same primary category.
Trusted references
FAQ
What is affected by CVE-2026-64650?
vercel @ai-sdk/harness-codex versions listed as affected should be reviewed: < 1.0.29.
What should I fix first?
Start with internet-facing sites, admin panels, login flows, plugins, themes, modules, packages, and systems that process user-controlled input or sensitive data.
How do I confirm the fix worked?
Apply the patched version or mitigation, clear caches where relevant, retest the affected workflow, and run a new Fixnx scan to verify public website exposure signals.
How are Fixnx security risk categories chosen?
Fixnx keeps one canonical risk page and assigns only broad, relevant categories such as ecosystem, technology area, or vulnerability class.