cac-windows
Claude Code Windows privacy tool — binary patches to remove Chinese identity markers.
This report shows the evidence for cac-windows@1.0.26. 5 of 9 known versions are blocked or flagged. See all versions.
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThis is a concrete automatic lifecycle chain that modifies and redirects a foreign AI-agent CLI. The observed behavior meets the install-hook abuse block boundary even without proof of secret theft.
Decision evidence
public snapshot- A postinstall hook runs automatically on package installation.
- On Windows, the hook installs a pinned Claude Code release from the npm registry without an explicit command from the user.
- The hook invokes the foreign package's install script and then runs a patcher against its executable.
- The patcher changes bytes in Claude Code binaries and saves a backup.
- The shipped executable also changes Claude identity values and installs telemetry-blocking interception code.
- The tool gives its wrapper precedence over the normal Claude command through the user PATH.
- No source evidence of secret collection or exfiltration to an attacker-controlled endpoint was found.
- The only install-time network host observed is the npm registry.
Source & flagged code
6 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgA postinstall hook runs automatically on package installation.
package.jsonView on unpkg · L20Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to cac-windows@1.0.26.
The versions list below shows the verdict for each recorded version.
- Avoid installing cac-windows@1.0.26. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.