koragraphmcp@0.2.0
Local multi-repo code-graph MCP server
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe source establishes an unconsented install-time project mutation and user-command-driven agent and Git hook persistence. It does not establish the narrower concrete malicious chain required for a publish block.
Decision evidence
public snapshot- The automatic postinstall hook writes KORAINIT.md into a consumer project without an explicit setup command.
- The installed onboarding file instructs an agent to create or refresh agent instruction files for several vendors.
- The ingest command automatically installs Claude Code command hooks and Git hooks unless the user opts out.
- The hook installer writes command entries to the target repository's .claude/settings.json.
- The postinstall hook only copies a local Markdown file and exits for global installs.
- No credential harvesting, outbound network client, remote payload download, or self-dependency was found in the inspected paths.
- Agent and Git hook changes occur through package commands, not directly during postinstall.
Source & flagged code
15 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgA bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to koragraphmcp@0.2.0.
See all versions of koragraphmcp for other recorded verdicts.
- Review the evidence and your use of koragraphmcp@0.2.0 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.