@clawos-dev/clawd@0.2.641
Standalone clawd daemon — Claude Code (and future Codex) session server over WebSocket
AI Security Review
scanned 11d ago · by lpm-firewall-aiSource inspection found no install lifecycle hook or concrete malicious behavior. The dynamic-code findings occur in generated validation and bundled compatibility code; project operations are exposed through explicit package tools.
Decision evidence
public snapshot- package.json exposes a CLI and has no install-time lifecycle hook.
- The app-builder bundle maps its publish workflow to a package-owned project collector; its removal handler runs only when that tool is called for a named project.
- The CLI appends a turn-duration record to the matching Claude session log, an integration behavior rather than credential harvesting.
- The flagged MCP bundles contain runtime bridge and daemon configuration paths; inspection found no concrete malicious action in those paths.
- The app-builder tool can invoke a removal script and delete a project directory after the tool is called.
- The app-builder bundle uses dynamic function construction for generated validation code and a bundled compatibility helper.
Source & flagged code
15 flagged · loading sourcePackage source references child process execution.
dist/app-builder-plugin/mcp-server.cjsView on unpkg · L30858Package source references dynamic code evaluation.
dist/app-builder-plugin/mcp-server.cjsView on unpkg · L9420Source file is highly similar to a previously finalized malicious package; route for source-aware review.
dist/app-builder-plugin/mcp-server.cjsFindings
Affected versions and remediation
This report applies to @clawos-dev/clawd@0.2.641.
See all versions of @clawos-dev/clawd for other recorded verdicts.
- Review the evidence and your use of @clawos-dev/clawd@0.2.641 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.