@eventmodelers/cli@1.0.159
Eventmodelers CLI — real-time Claude agent + skills for Claude Code, for any stack (Node, Supabase, Axon, OpenCQRS, UmaDB, Kurrent, or modeling-only)
AI Security Review
scanned 17d ago · by lpm-firewall-aiSource inspection found no malicious install behavior. The explicitly started code export server does expose an unauthenticated deletion endpoint with wildcard CORS, creating a real but bounded destructive capability, so warn rather than block.
Decision evidence
public snapshot- The manifest declares no install lifecycle hook, and the package does not list itself as a dependency.
- The listener starts through the explicit `eventmodelers listen` command and runs the code export server.
- The server uses wildcard CORS and calls `listen` without specifying a host, with port 3001 as the default.
- The POST `/api/delete-slice` handler accepts a slice ID without an authentication check and deletes a matching `code-slice.json` under `.slices`.
- The deletion is scoped to a matching slice file; source does not show deletion outside `.slices`.
- The separate realtime code subscribes to configured topics and handles events in the user-started modeling loop; its retry logic only re-subscribes.
- The vulnerable listener requires a user to start it; the manifest has no install hook that starts it automatically.
Source & flagged code
9 flaggedPackage source references child process execution.
stacks/bridge/templates/bridge/ralph-local-ai.jsView on unpkg · L12Package source references shell execution.
stacks/bridge/templates/bridge/ralph-hook.jsView on unpkg · L68Findings
Affected versions and remediation
This report applies to @eventmodelers/cli@1.0.159.
See all versions of @eventmodelers/cli for other recorded verdicts.
- Review the evidence and your use of @eventmodelers/cli@1.0.159 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.