Loading vulnerability…
CVE-2025-54416 is a command injection vulnerability in the tj-actions/branch-names GitHub Action. The issue affects versions 8.2.1 and earlier and stems from inconsistent input sanitization and unescaped output handling when the action retrieves and exposes branch or tag names. An attacker can create a specially crafted branch name or tag containing shell metacharacters; when a downstream workflow consumes that output in a shell context, the injected content can be interpreted as commands. The vulnerable condition is therefore in the action’s handling of untrusted branch/tag name data and its propagation into downstream workflow steps.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This repository is a minimal proof-of-concept demonstrating CVE-2025-54416 in tj-actions/branch-names. It contains two files: a GitHub Actions workflow (.github/workflows/vulnerable-demo.yml) and a README describing the vulnerability and attack chain. The workflow is the core exploit artifact. It shows two jobs, build and deploy, both using tj-actions/branch-names@v8.2.1 to obtain the current branch name and then unsafely interpolating that attacker-controlled value into shell commands and Docker/AWS-related operations. The demonstrated capability is command injection in the GitHub Actions runner context via malicious branch names, with likely outcomes including secret exfiltration (GITHUB_TOKEN, AWS credentials), unauthorized Docker image build/push/pull actions, and CI/CD compromise. The repository does not include an automated exploit script beyond the vulnerable workflow itself, so it is best classified as a PoC rather than an operational exploit framework. The main fingerprintable observables are the vulnerable action reference, example attacker exfiltration domains/URLs, GitHub Actions environment/output files, AWS region/configuration, and example sensitive file access path.
5 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.