Skip to main content
Mallory
HighPublic exploitVulnerability

CVE-2026-71963: GitSpawn remote code execution in Hermes Agent

Published Sep 3, 2026 · Updated 23 days ago
Trending this weekNot trendingOutside this week’s top 100
Mention activity · 30 daysNo mention activity in the last 30 days.
CVSS8.6High · v4
EPSS0.86%59th percentile
CISA KEVNoNot listed
In the wildNonePublic exploit, no exploitation reported
IdentifiersCVE-2026-71963CWE-78 · Improper Neutralization of Special…

CVE-2026-71963 is a remote code execution vulnerability affecting Hermes Agent versions 0.18.2 through 0.21.0. During repository-context gathering, the agent invokes git status without neutralizing repository-controlled Git configuration. An attacker can supply a repository whose core.fsmonitor setting specifies a malicious command. When a user opens the repository and sends any message, Git refreshes its index and executes the configured command with the user's privileges, outside the agent's sandbox and approval controls. The vulnerability is part of the GitSpawn vulnerability class.

Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Analyst brief

Impact, mitigation & remediation

What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation compromises confidentiality, integrity, and availability within the affected user's security context. Attacker-controlled commands can access the agent's environment, including configured provider API keys, read or modify accessible files and repositories, and disrupt user-accessible resources. Agent sandboxing and command-approval controls do not contain this execution path.

Mitigation

If you can’t patch tonight, do this now.

Until patched, avoid opening or interacting with untrusted repositories in affected Hermes Agent versions. Inspect repository-local .git/config files for command-executing settings, particularly core.fsmonitor, before using the agent. For the specific filesystem-monitor execution path, override repository configuration on every background Git invocation, for example with git -c core.fsmonitor=false status. Minimize sensitive environment variables and API keys available to the agent. Disabling core.fsmonitor alone does not address other potentially command-executing Git settings.

Remediation

Patch, then assume compromise.

Update Hermes Agent to a build containing fix commit f6234d00c5d59450adea1d7edd30ad3859375c79. Versions 0.18.2 through 0.21.0 are affected; a specific fixed release number is not available.

FAQ

Frequently asked questions

Is CVE-2026-71963 exploited in the wild?
As of Oct 25, 2026, Mallory has no reports of CVE-2026-71963 being exploited in the wild. It is not listed in CISA KEV.
Is there a public exploit for CVE-2026-71963?
Yes. Mallory tracks 1 public exploit or proof of concept for CVE-2026-71963, the earliest disclosed Sep 27, 2026.
Is there a patch for CVE-2026-71963?
Update Hermes Agent to a build containing fix commit f6234d00c5d59450adea1d7edd30ad3859375c79. Versions 0.18.2 through 0.21.0 are affected; a specific fixed release number is not available.
How severe is CVE-2026-71963?
CVE-2026-71963 has a CVSS v4 score of 8.6 (High) and an EPSS score of 0.86% (59th percentile) as of Oct 25, 2026.
Activity feed

Recent activity

8 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

View more in app →
cysecurity newsNews · Sep 4, 20261 Folder Was All It Took: Security Researchers Find AI Coding Agents Can Be Hijacked Before a Single Prompt Is Typed - CySecurity News - Latest Information Security and Hacking IncidentsA GitSpawn-related arbitrary code-execution vulnerability affecting Nous Research's Hermes Agent. Opening a maliciously supplied repository containing a .git configuration that invokes an external Git helper can cause the agent's automatic background Git commands to execute attacker-controlled code as the local developer.Read more ↗ghsa syntheticNews · Sep 3, 2026GHSA-cc88-9pxf-j2wvA remote code execution vulnerability in Hermes Agent versions 0.18.2 through 0.21.0. A malicious repository can set Git's core.fsmonitor configuration to an attacker-controlled command; when a user opens that repository and sends a message, Hermes Agent triggers a git status index refresh that executes the command in the user's process context, potentially exposing environment variables including provider API keys.Read more ↗cvefeed high severityNews · Sep 3, 2026CVE-2026-71963 - Hermes Agent 0.18.2 - 0.21.0 RCE via git core.fsmonitor Config InjectionA remote code execution vulnerability in Hermes Agent versions 0.18.2 through 0.21.0. A malicious repository can configure Git's core.fsmonitor to an attacker-controlled command; when a user opens that repository and sends a message, Hermes Agent triggers a git status index refresh that executes the command in the user's process context, potentially exposing environment variables including configured provider API keys.Read more ↗cyber security newsNews · Sep 2, 2026GitSpawn Flaws Let Malicious Repositories Execute Code in Claude Code, Codex, Cursor and GrokA vulnerability in Hermes Agent allows arbitrary command execution through malicious repository configuration during startup context gathering. Background Git commands trigger the repository's core.fsmonitor helper with the user's privileges, outside agent sandbox and permission controls. The repository must arrive as files with its .git directory intact. The vendor reportedly did not respond to six contact attempts; patch availability is not explicitly established.Read more ↗
+2 more in the timeline. Sign up to see them
Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Mallory platform

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, and which detections to deploy tonight.
Exposure mappingQuery your assets running an affected version, and investigate the blast radius.
Threat actor evidenceEvery report linking this CVE to a named adversary.
Associated malwareMalware families linked to this CVE, with evidence and IOCs.
Detection signaturesYARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Vendor-by-vendor mappingCross-references every affected SKU, including bundled OEM variants.
Social activity2Community discussion across Reddit, Mastodon, and other social sources.