@whalent/agent-core@0.3.458
Core runtime for Whalent Agent
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe published manifest has no install lifecycle and no self-dependency, and the readable behavior matches a user-launched agent core that talks to its own providers and a local Jupyter sidecar. Obfuscation and process, SSH, and API-key handling are package-aligned capabilities without a concrete exfiltration or control-hijack chain.
Decision evidence
public snapshot- Published package.json is @whalent/agent-core 0.3.458 with main and bin dist/index.cjs and no preinstall, install, or postinstall scripts.
- Registry dependencies are only ssh2 and optional node-pty; the published manifest does not depend on its own package name.
- dist/mcp.cjs is a Node entry that freezes provider auth environment key names such as ANTHROPIC_API_KEY for the agent client.
- dist/jupyter_sidecar.py is a local stdin/stdout NDJSON Jupyter kernel helper owned by this agent runtime.
- dist/index.cjs and dist/mcp.cjs are javascript-obfuscator style bundles, so large parts of the runtime are not plain source.
- Readable names include child_process, writeDesktopLink under a Whalent path, ssh2, and provider API key handling consistent with a user-launched agent.
Source & flagged code
7 flagged · loading sourceSource contains an obfuscated payload loader that reconstructs and executes hidden code.
dist/mcp.cjsView on unpkg · L1A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
dist/mcp.cjsView on unpkgFindings
Affected versions and remediation
This report applies to @whalent/agent-core@0.3.458.
See all versions of @whalent/agent-core for other recorded verdicts.
- Review the evidence and your use of @whalent/agent-core@0.3.458 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.