GhostAction: Tens of thousands of repositories GitHub compromised with malware that steals credentials of cloud and AI
During the weekend of November 7-9, 2026, a new wave of the attack GhostAction was detected.
Tens of thousands of repositories GitHub were infiltrated with workflows malicious.
The goal: to steal credentials from AWS, Azure, OpenAI, and any cloud service or AI developers to have in your code.
If you’re a developer and you’re using GitHub, your repo could be committed. Now.
What happened
The attack
FACT (according to GB Hackers):
A new wave of the attack GhostAction is pushing GitHub Actions workflows malicious repositories.
FACT: The range is tens of thousands of repositories infiltrates.
FACT: The malware is looking for:
- Credentials of cloud (AWS keys, Azure tokens, GCP credentials)
- Credentials IA (OpenAI API keys, Anthropic tokens, etc)
- In the source code
- In the git history complete (commit earlier)
How it works
- Workflow is malicious pushed to the repository (via engagement of the repo or the account of the owner)
- The Workflow runs automatically on GitHub Actions
- You search for files containing credentials (env files, config files, code)
- Looking at the whole history of your git credentials that were removed but they were in history
- Exfiltría any key/token/credential found
Why is critical
Vector 1: Credentials of cloud exposed
If your AWS key was stolen, the attacker has access to:
- All your servers
- All your databases
- All of your backups
- Potentially your customer data
Cost of remediation: Thousands to millions of dollars + downtime + gap of privacy.
Vector 2: Credentials IA = access to model + expenses
If your OpenAI API key was stolen, the attacker:
- You can use your API key to run queries in GPT-4
- The expenses fall in to your account
- You can use your quota before you know it
- Access to your information (if you used the API with sensitive data)
Cost: Potentially thousands of dollars in queries is not authorized in minutes.
Vector 3: Supply chain amplification
If an open source developer is committed:
- Your repo is infiltrated
- Workflow malicious run in your CI/CD
- Credentials of the developer stolen
- But also: potentially, the credentials of any service that you use in your development (secrets management, deployment, etc)
A developer committed = potentially hundreds of projects dependent on risk.
The pattern of attack
Step 1: Access to the repository
The attacker gains access via:
- Credentials compromised from the developer
- GitHub account hacked
- Access Token to the personal access token is stolen
Step 2: Inject workflow malicious
In .github/workflows/, the attacker creates or modifies a workflow YAML that:
- Is run on each push/PR (normal)
- Run command to extract credentials
- Exfiltría to a server controlled by the attacker
Step 3: auto-Run
Every time someone pushea code:
- The workflow runs automatically
- Credentials are extracted
- Data are exfiltrados
The developer does not notice anything (workflow seems legitimate).
How to 24Cevent would have helped
If the developers/companies have detection of anomalies in its pipeline CI/CD:
Detection of workflow anomalous
When the malware is injected:
- YAML file: New workflow was added with no change in repo ‘normal’
- Correlation: the Change coincides with remote access anomalous GitHub
- Action: Blocking of merge + research
The malware is never mergeado.
Detection of exfiltration of credentials
When the workflow tries to exfiltration:
- Network: Connection to unknown IP from CI/CD environment
- Data: Volume anomalous data coming out of the CI/CD
- Correlation: Data being exfiltrados to unknown location
24Cevent: ‘CREDENTIAL EXFILTRATION DETECTED’ + lock immediately.
Detection of multiple repositories committed
When GhostAction is sneaking up tens of thousands of repos simultaneously:
- Correlation: the Same pattern of attack in 1000+ repos simultaneously
- Analysis: Supply chain attack in progress
- Escalation: ‘GITHUB-WIDE SUPPLY CHAIN ATTACK DETECTED’
GitHub + developers notified immediately.
The lesson
This attack demonstrates that:
- The developers are targets of high value (credential = access to everything)
- CI/CD pipelines are vectors of attack (malicious code is executed automatically)
- Git history is a risk (credentials removed are still in history)
For developers:
- Use secrets management (not in code)
- Rotate credentials regularly
- Audit git history for credentials
- Monitor CI/CD for changes anomalous
SOURCES CHECKED: