openkrak-mcp@1.0.7
OpenKrak MCP Server - AI coding intelligence via Dorchester engine
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe package has no automatic install execution, but it contains undisclosed remote telemetry and AI-agent steering instructions in its runtime tool definitions. These create a concrete privacy and agent-integrity risk without proving broader malware behaviour.
Decision evidence
public snapshot- Runtime posts a host-derived fingerprint and usage metrics to a remote service.
- Runtime posts caught error text, which can contain local path or analysis details, to the same remote service.
- The MCP tool descriptions instruct an agent not to read source files, attempting to constrain independent inspection.
- The README says there is no telemetry, but the bundled runtime implements telemetry.
- The manifest has only a publish-time build script and no install lifecycle hook.
- Network requests occur only after the configured MCP server receives a tool call.
- No source evidence shows shell execution, persistence, project-file mutation, or self-dependency.
Source & flagged code
4 flagged · loading sourceSource collects local host identity data and sends it to an external endpoint.
dist/license/check.jsView on unpkg · L2Source fingerprint signature matches a known malicious package signature; route for source-aware review.
dist/license/check.jsView on unpkgFindings
Affected versions and remediation
This report applies to openkrak-mcp@1.0.7.
See all versions of openkrak-mcp for other recorded verdicts.
- Review the evidence and your use of openkrak-mcp@1.0.7 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.