Claude Skill Antivirus Local Skill Directory Executable File Bypass

Executive Summary

CVE-2026-84810 is a flaw in claude-skill-antivirus versions 2.1.3 and earlier where the tool fails to analyze executable files during scans. It only checks the SKILL.md manifest file and ignores Python source code, bytecode, and other artifacts in the scripts directory. This allows attackers to embed malicious code in non-manifest files that receive a SAFE verdict despite containing unanalyzed executable payloads.

Useful 0 Not Useful 0

Detection Guidance

To detect this vulnerability, inspect local skill directories for executable files in scripts/ or __pycache__/ that are not referenced in SKILL.md. Check if malicious Python bytecode (.pyc) or scripts execute during import without triggering antivirus scans. Review downloader.js for improper file loading in fetchLocal().

Commands: 1) Find unreferenced executables: find /path/to/skills -type f \( -name "*.pyc" -o -name "*.py" \) -exec grep -L "SKILL.md" {} \; 2) Test import behavior: python3 -c "import malicious_skill; print('Executed')" 3) Audit downloader.js for fetchLocal() implementation.

Useful 0 Not Useful 0

Impact Analysis

This vulnerability allows attackers to bypass antivirus detection by hiding malicious code in files like Python scripts or bytecode. The antivirus would still rate the skill as SAFE with a 100/100 trust score, potentially leading to execution of harmful code on your system when the skill is imported or used.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability could lead to non-compliance with GDPR and HIPAA due to the failure to detect malicious executable files in skills. GDPR requires protection of personal data, and HIPAA mandates safeguards for health information. If malicious code bypasses antivirus scans, it may exfiltrate sensitive data, violating these regulations.

Useful 0 Not Useful 0

Mitigation Strategies

Immediately upgrade to a patched version of claude-skill-antivirus (v2.1.4 or later). If unavailable, modify the scanner to analyze all files in skill directories, not just SKILL.md. Temporarily disable local skill directory scanning until fixed. Monitor for suspicious activity in scripts/ or __pycache__.

As a workaround, manually inspect all files in skill directories for malicious code. Block untrusted skill installations. Report any bypass attempts to the vendor. The root cause requires changing the file-loading architecture in downloader.js to scan all artifacts.

Useful 0 Not Useful 0