claude-multiacc@2.0.59
Unified Claude Code and OpenAI Codex subscription pooling with quota-aware selection.
AI Security Review
scanned 3d ago · by lpm-firewall-aiSource inspection resolves the lifecycle hook as guarded product setup, and the scanner-named test path contains ordinary test code. The account token requests and update notifier show no confirmed unrelated credential transfer or other attack behavior.
Decision evidence
public snapshot- package.json defines a postinstall hook; scripts/postinstall.mjs limits automatic setup to global, non-CI installs and delegates to the package installer.
- install.sh creates package-labeled account jobs as part of the add-on setup. The named test file contains a test fixture, not an executable payload.
- The account tools send bearer tokens in usage requests; inspected callers are account-management paths, and no unrelated credential recipient was identified.
- The selector reads a JSON request from stdin and passes it to its selection function; the Claude shim ultimately invokes the discovered real CLI with the caller's arguments.
- The prepared update-notifier manifest has no install hook; its entrypoint constructs the notifier and checks for updates.
Source & flagged code
7 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgpackage.json defines a postinstall hook; scripts/postinstall.mjs limits automatic setup to global, non-CI installs and delegates to the package installer.
package.jsonView on unpkg · L57Findings
Affected versions and remediation
This report applies to claude-multiacc@2.0.59.
See all versions of claude-multiacc for other recorded verdicts.
- Review the evidence and your use of claude-multiacc@2.0.59 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.