@twing/cli@1.1.0
AI Security Review
scanned 11d ago · by lpm-firewall-aiInspected source shows an install-time machine setup gate and agent integration, but no concrete malicious behavior. The findings are consistent with guarded setup of this package’s own hooks, so mark clean.
Decision evidence
public snapshot- package.json runs `node postinstall.cjs`; the hook exits unless npm reports a global install or an explicit machine-setup flag.
- Machine setup wires Claude Code and OpenCode settings and writes an auto-managed marker; this is broad machine-level agent integration.
- The flagged target files describe committed repo hooks, explicit `twing init`, and primary-working-directory resolution; their cited passages are descriptive comments.
- The adapter resolves and runs a configured Twing hook, while test-support’s child process runs `git init` in a temporary test repository.
- No concrete credential theft, unrelated recipient, destructive action, or concealed payload execution was established from the inspected source.
- The postinstall gate excludes ordinary project installs and self-installs unless the machine-setup flag is passed.
- The inspected adapter, test-support, and SQLite-source excerpts do not establish hostile execution; machine integration appears to install Twing’s own hook/plugin.
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 source references child process execution.
dist/opencode-adapter.jsView on unpkg · L12Findings
Affected versions and remediation
This report applies to @twing/cli@1.1.0.
See all versions of @twing/cli for other recorded verdicts.
- Review the evidence and your use of @twing/cli@1.1.0 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.