Microsoft discovers prompt injection vulnerability in Claude Code GitHub Action exposing CI/CD secrets
By
Microsoft Defender Security Research Team, Dor Edry, Amit Eliahu
Summary
Microsoft Threat Intelligence discovered a prompt injection vulnerability in Anthropic's Claude Code GitHub Action that could expose CI/CD workflow secrets. The vulnerability allowed AI agents processing untrusted GitHub content (issue bodies, PR descriptions, comments) to access /proc/self/environ via the Read tool, which was not sandboxed like subprocess execution paths. This could leak the ANTHROPIC_API_KEY and other credentials available to the runner. The article covers the attack chain, responsible disclosure process, Anthropic's mitigation, and guidance for securing AI-powered CI/CD workflows.
Source
bskyMicrosoft discovers prompt injection vulnerability in Claude Code GitHub Action exposing CI/CD secretsmicrosoft.comKey quotes
· 3 pulledMicrosoft Threat Intelligence discovered that Anthropic's Claude Code GitHub Action could expose CI/CD workflow secrets when AI agents process untrusted GitHub content, including issue bodies, pull request descriptions, and comments.
While Claude Code Action supported environment scrubbing for subprocess execution paths such as Bash, the Read tool was not subject to the same sandboxing model.
It was eventually authorized to access /proc/self/environ, reading the workflow's ANTHROPIC_API_KEY and potentially other credentials available to the runner.
You might also wanna read
Security Vulnerability in Claude Cowork Enables File Exfiltration via Prompt Injection
The article reveals a security vulnerability in Anthropic's Claude Cowork feature, demonstrating how it can be exploited for file exfiltrati
promptarmor.com·5mo ago
Anthropic Releases Claude Code Security AI Tool to Help Defenders Detect Vulnerabilities
Anthropic is releasing Claude Code Security, an AI-powered cybersecurity tool designed to help defenders detect novel, high-severity vulnera
anthropic.com·4mo ago
Anthropic's Claude Opus 4.6 AI Model Discovers 500+ High-Severity Security Flaws in Open-Source Libraries
Anthropic's latest AI model, Claude Opus 4.6, has discovered over 500 previously unknown high-severity security vulnerabilities in open-sour
axios.com·4mo ago
Analysis: Anthropic Blocks "OpenCode" Term in Claude Code OAuth System Prompts
This article presents evidence that Anthropic is specifically blocking the term "OpenCode" in Claude Code's OAuth system prompts. The conten
gist.github.com·5mo ago
Analysis: Anthropic's Claude Mythos Marketing Uses Previously Discovered FreeBSD Vulnerability
The article criticizes Anthropic's marketing of its Claude Mythos Preview AI model, which is being showcased using CVE-2026-4747 - a 17-year
flyingpenguin.com·2mo ago
Anthropic Launches Safer 'Auto Mode' for Claude Code AI Tool
Anthropic has introduced an 'auto mode' for Claude Code, their AI tool that can make permissions-level decisions on users' behalf. This new
The Verge·3mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.