CVE-2026-64650

MEDIUMPre-NVD 6.36.3
EchelonGraph scoreMEDIUM confidence

This medium-severity CVE scores 6.3 under a secondary CVSS source (NVD's own analysis pending). EPSS exploit probability: 0.1%, top 98% of all CVEs by exploit prediction. GitHub Security Advisory data not yet ingested — confidence will rise once GHSA publishes (typical lag: hours to days for open-source ecosystem CVEs; never for infrastructure-only CVEs).

Triggered by: NVD CVSS baseline
Sources: epss, secondary
6.3EG
EchelonGraph verdictMonitorLow exploitation likelihood right now — keep watching.
  • Lower severity and no public exploit yet
CISA-KEV: Not listedEPSS PROB: 0%CVSS: 6.3Exploit: None knownExposed: 0

No vendor fix yet — apply a workaround or compensating control (WAF / firewall / segmentation) and watch for a patch.

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.

CVSS v3
6.3
EG Score
6.3(medium)
EG Risk
28(Track)
EG Risk 28/100SSVC: Track

EG Risk is EchelonGraph's 0–100 priority score: it fuses intrinsic severity with real-world exploitation and automatability so you can rank equal-severity CVEs and fix the most dangerous first. Higher = act sooner. Distinct from the 0–10 EG Score (severity).

How it’s computed
Severity63% × 45%
Exploitation0% × 40%
Automatability0% × 15%
Action: Routine — remediate on your standard cadence.
EPSS PROB
0%
EPSS %ILE
2%
KEV
Not listed

Published

July 20, 2026

Last Modified

July 23, 2026

Advisory Details (2)

Auto-updated Jul 20, 2026
Patch available. Sources: github, github_pr.
github_pr Patch Available

fix(harness): improve CLI relay tool invocation logic

Patch available: vercel/ai [email protected] (PR #17105 merged 2026-07-10)

https://github.com/vercel/ai/pull/17105
github Patch Available

AI SDK Codex Harness Tool Relay Authorization Bypass · Advisory · vercel/ai · GitHub

https://github.com/vercel/ai/security/advisories/GHSA-qw9h-448j-6rph

Weakness Classification(1)

MITRE Common Weakness Enumeration — the root-cause categories this CVE belongs to.

Data Freshness Timeline

(refreshed 14× in last 7d / 24× in last 30d)

Each row is a source pipeline that fetched or updated this CVE on that date, with what changed. For example, "NVD update" means NVD published or revised its analysis for this CVE; "MITRE cvelistV5" means we ingested or refreshed it from the CNA feed. Most recent first.

  1. 2026-07-30 16:28 UTCEPSS rescore
  2. 2026-07-30 12:02 UTCEG score recompute
  3. 2026-07-30 01:30 UTCEPSS rescore
  4. 2026-07-30 01:30 UTCEPSS rescore
  5. 2026-07-29 10:08 UTCEG score recompute
  6. 2026-07-28 15:37 UTCEPSS rescore
  7. 2026-07-27 06:23 UTCEG score recompute
  8. 2026-07-26 14:54 UTCEPSS rescore
  9. 2026-07-26 14:54 UTCEPSS rescore
  10. 2026-07-26 04:26 UTCEG score recompute
  11. 2026-07-25 14:18 UTCEPSS rescore
  12. 2026-07-25 14:18 UTCEPSS rescore
  13. 2026-07-24 14:18 UTCEPSS rescore
  14. 2026-07-24 14:18 UTCEPSS rescore
  15. 2026-07-23 14:18 UTCEPSS rescore
  16. 2026-07-23 14:18 UTCEPSS rescore
  17. 2026-07-23 03:22 UTCEG score recompute
  18. 2026-07-22 14:08 UTCEPSS rescore
  19. 2026-07-22 14:08 UTCEPSS rescore
  20. 2026-07-21 15:25 UTCEPSS rescore
  21. 2026-07-21 15:25 UTCEPSS rescore
  22. 2026-07-20 22:42 UTCEG score recompute
  23. 2026-07-20 20:53 UTCEG score recompute
  24. 2026-07-20 20:52 UTCMITRE cvelistV5first tracked

Frequently asked(5)

What is CVE-2026-64650?
CVE-2026-64650 is a medium vulnerability published on July 20, 2026. 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…
When was CVE-2026-64650 disclosed?
CVE-2026-64650 was first published in the National Vulnerability Database on July 20, 2026, with the most recent update on July 23, 2026. EchelonGraph re-ingests CVE updates from NVD on a 2-hour cycle, so this page reflects the latest published state.
Is CVE-2026-64650 actively exploited?
CVE-2026-64650 is not currently on CISA's Known Exploited Vulnerabilities catalog. FIRST EPSS estimates a 0% probability of exploitation in the next 30 days, which ranks it in the top 98.4% of all scored CVEs.
What is the CVSS score of CVE-2026-64650?
CVE-2026-64650 has a CVSS v3 base score of 6.3 (NVD).
How do I remediate CVE-2026-64650?
Patch to the fixed version published by the affected vendor. Where vendor advisories exist for CVE-2026-64650, EchelonGraph cross-links them in the Vendor Advisories panel below — those typically contain the canonical remediation steps, fixed version numbers, and any vendor-specific mitigations.

Dependency Blast Radius

Explore the affected products and dependency analysis for CVE-2026-64650

Explore →

Is Your Infrastructure Affected by CVE-2026-64650?

EchelonGraph automatically scans your cloud infrastructure and maps CVE exposure using blast radius analysis.