@velaro/cli@1.4.47
Velaro Workspace v20 — command-line interface for managing bots, knowledge base ingestion, MCP API keys, search indexes, and ops health.
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe install hook is limited to a banner, but the package contains an explicit command that modifies a broad AI-agent control surface and installs a runtime MCP dependency through npx. This is a real lifecycle risk requiring a warning, not evidence of automatic malicious installation behavior.
Decision evidence
public snapshot- The package has an install-time postinstall hook.
- The explicit mcp-key install command writes a persistent Claude Code MCP configuration.
- That configuration starts npx with @velaro/mcp-server, allowing a later agent startup to fetch and execute that separate package.
- The KB screenshot command can retrieve service credentials and upload captured content, but only after the user invokes that command.
- The postinstall hook only imports and displays the package banner; no network or shell action is shown there.
- The Claude configuration mutation requires the named user command and creates a vendor-named entry.
- The update command's global npm installation is also only reachable through the named update command.
- No hidden payload, credential exfiltration path, or automatic foreign agent-config mutation was found.
Source & flagged code
9 flaggedPackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgPackage manifest contains a dependency pinned to a remote tarball URL.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to @velaro/cli@1.4.47.
See all versions of @velaro/cli for other recorded verdicts.
- Review the evidence and your use of @velaro/cli@1.4.47 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.