karajan-code@4.58.0
Local multi-agent coding orchestrator with TDD, SonarQube, and code review pipeline
AI Security Review
scanned 20h ago · by lpm-firewall-aiThe automatic global agent configuration change is a real consent risk, but it registers this package’s own MCP server rather than showing a concrete attack. Classify as suspicious and warn for agent extension lifecycle risk.
Decision evidence
public snapshot- package.json runs scripts/postinstall.js, whose main function unconditionally configures both Claude Code and Codex.
- The postinstall writes a karajan-mcp entry to the global Claude and Codex configuration, pointing to this package’s bundled server; it does so without asking for consent.
- The bundled server identifies itself as karajan-mcp and uses the MCP server interface. A separate installer asks whether to configure MCP, while `kj init` exposes an opt-out for the ai-trash hook.
- Other flagged locations implement a settings patcher, an installer helper, a test fixture, configuration inventory or validation, template setup, or project-document discovery; they do not establish another install-time agent mutation.
- The inspected registration points to this package’s own MCP server over standard input and output; the setup path does not show an unrelated recipient or credential transfer.
Source & flagged code
24 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.
scripts/dual-publish.mjsView on unpkg · L17Findings
Affected versions and remediation
This report applies to karajan-code@4.58.0.
See all versions of karajan-code for other recorded verdicts.
- Review the evidence and your use of karajan-code@4.58.0 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.