# CVE-2026-47751: Anthropic Claude Code Action arbitrary code execution in GitHub Actions runner

Source: https://junglewise.ai/threats/cve-2026-47751-anthropic-claude-code-action-arbitrary-code-execution-in-github  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-07-16. Updated: 2026-07-16  
License: CC BY 4.0 with a link back

CVE-2026-47751 · Severity: medium · CVSS 5.3 · Published 2026-07-16

Anthropic's Claude Code Action is a tool used to automate code reviews and issue management on GitHub. A security flaw allowed attackers to execute malicious code on the GitHub infrastructure by submitting a specially crafted pull request. This could lead to the theft of sensitive credentials, such as API keys and tokens, stored within the GitHub environment.

Vendors: [Anthropic](https://junglewise.ai/threats/vendors/anthropic), [GitHub Actions](https://junglewise.ai/threats/vendors/actions).

## Quick answers

**Is CVE-2026-47751 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-47751 being exploited in the wild.

**How severe is CVE-2026-47751?** Medium severity, with a CVSS score of 5.3, published 16 July 2026.

**Which products and versions are affected?** anthropics/claude-code-action (GitHub Actions) < 1.0.74.

**Is there a fix for CVE-2026-47751?** A patch was published on 2026-03-17: Fix merged into main branch Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-47751 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-47751?** Check whether you run Anthropic Claude Code Action or anthropics/claude-code-action (GitHub Actions) at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

Claude Code Action prior to version 1.0.74 is vulnerable to arbitrary code execution and secret exfiltration. The vulnerability exists because the action checks out the head branch of a pull request (which is attacker-controlled) and subsequently reads configuration files like .mcp.json and .claude/settings.json from the working directory. Because the action unconditionally enabled all project Model Context Protocol (MCP) servers and executed hooks or shell commands defined in these files before tool-permission gating, an attacker could achieve RCE on the GitHub Actions runner. This allows for the exfiltration of workflow secrets. The fix in version 1.0.74 ensures that sensitive configuration files are restored from the trusted base branch before the CLI execution.

## Affected products

- Anthropic claude-code-action < 1.0.74

## Identifiers

- CVE-2026-47751
- GHSA-8q5r-mmjf-575q

## Timeline

- 2026-03-17: patched: Fix merged into main branch
- 2026-03-18: advisory: Version 1.0.74 released
- 2026-05-20: disclosed: GitHub security advisory published
- 2026-07-16: advisory: NVD published CVE-2026-47751

## References

- [https://github.com/anthropics/claude-code-action/commit/9ddce40de8c1ab71fb6303a125fdad0968dc1312](https://github.com/anthropics/claude-code-action/commit/9ddce40de8c1ab71fb6303a125fdad0968dc1312)
- [https://github.com/anthropics/claude-code-action/pull/1066](https://github.com/anthropics/claude-code-action/pull/1066)
- [https://github.com/anthropics/claude-code-action/releases/tag/v1.0.74](https://github.com/anthropics/claude-code-action/releases/tag/v1.0.74)
- [https://github.com/anthropics/claude-code-action/security/advisories/GHSA-8q5r-mmjf-575q](https://github.com/anthropics/claude-code-action/security/advisories/GHSA-8q5r-mmjf-575q)