CVE-2026-64650 vercel CVE debrief

Who should care

Defenders responsible for managing the `@ai-sdk/harness-opencode` tool, particularly those working with untrusted code in sandboxed environments, should assess their exposure and take necessary actions to mitigate the vulnerability.

Why it matters

The vulnerability in `@ai-sdk/harness-opencode` allows untrusted code to invoke host-exposed tools without authorization, posing a risk to systems using this tool. Defenders should update to version 1.0.29 or apply workarounds to mitigate this risk.

• Potential unauthorized invocation of host-exposed tools
• Possible secret lookups and deployment operations
• Cloud API calls without proper authorization
• Verification of authorization mechanisms for host-exposed tools

Technical summary

The `@ai-sdk/harness-opencode` tool, prior to version 1.0.29, has a vulnerability that allows untrusted code in the sandbox to invoke host-exposed tools without proper authorization. This is due to a process-path authorization fallback that was removed in version 1.0.29. The fix involves relay requests being accepted only after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event.

Defensive priority

Defenders should prioritize updating the `@ai-sdk/harness-opcode` tool to version 1.0.29 or later. In the meantime, restrict the execution of untrusted code in the sandbox and limit host-exposed tools to non-sensitive operations.

Recommended defensive actions

• Update `@ai-sdk/harness-opcode` to version 1.0.29 or later
• Restrict execution of untrusted code in the sandbox
• Limit host-exposed tools to non-sensitive operations
• Review compensating controls for exposed systems while remediation is scheduled and verified
• Check relevant monitoring, detection, and logs for exposed assets that need extra review
• Track exceptions, retest remediated assets, and close the item only after evidence is documented
• Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up

Evidence notes

The CVE record and NVD entry provide details on the vulnerability, including its description, CVSS score, and affected versions. References to GitHub pull requests and security advisories offer additional context.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.