nansen-cli@1.46.0
AI-agent CLI for Nansen API analytics, DEX swaps, and cross-chain trading
AI Security Review
scanned 1mo ago · by lpm-firewall-aiSource inspection supports a warning for conditional credential forwarding to an unrestricted custom HTTPS endpoint, not a finding of malware. Agent configuration changes and first-party skill onboarding are guarded or explicitly invoked and do not establish block-eligible install abuse.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
Invoke nansen mcp verify with --url and --send-api-key against a reachable custom MCP server.
Impact
The custom endpoint receives a Nansen credential that could permit unauthorized service access; explicit consent and transport checks limit activation but do not bind the key to Nansen.
Mechanism
The command resolves the saved or environment API key and sends an authenticated tools/call request to the selected endpoint without a service allowlist.
AI rationale
Source inspection supports a warning for conditional credential forwarding to an unrestricted custom HTTPS endpoint, not a finding of malware. Agent configuration changes and first-party skill onboarding are guarded or explicitly invoked and do not establish block-eligible install abuse.
Evidence
src/doctor.jssrc/commands/mcp.jssrc/mcp-verify.jssrc/cli.js
Network endpoints1
mcp.nansen.ai/ra/mcp
Decision evidence
public snapshotAI called this Suspicious at 97.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- Authentication resolves the Nansen API key from local configuration or NANSEN_API_KEY.
- The active MCP verify command accepts a custom URL and enables saved-key forwarding through --send-api-key.
- Custom HTTPS destinations have no service allowlist; saved-key forwarding is withheld unless explicitly authorized.
- The authenticated verification call sends the resolved key in the NANSEN-API-KEY header to the selected URL.
- Explicit MCP install commands merge the Nansen server into agent configuration and back up existing files.
- Global interactive postinstall onboarding offers installation of the first-party Nansen skill through a declineable prompt.
Evidence against
- No covert credential theft or malicious payload execution was identified.
- Verification rejects public plain HTTP destinations and refuses redirects.
- Local and noninteractive npm installations skip skill installation; MCP configuration changes require explicit CLI commands.
- The update checker fetches registry version metadata and caches it rather than executing downloaded code.
Behavioral surface
Source & flagged code
8 flaggedpackage.jsonView file
•scripts.postinstall = node scripts/postinstall.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/postinstall.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgsrc/wallet.jsView file
366patternName = generic_password
severity = medium
line = 366
Medium
Findings
4 High7 Medium4 Low
HighInstall Time Lifecycle Scriptspackage.json
HighSandbox Evasion Gated Capabilitysrc/doctor.js
HighSemantic Analysis Limitedsrc/x402.js#virtual:normalized:round1
HighPrevious Version Dangerous Deltasrc/update-check.js
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumSecret Patternsrc/wallet.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
MediumSecret Patternsrc/wallet.js
MediumSecret Patternsrc/wallet.js
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
Affected versions and remediation
This report applies to nansen-cli@1.46.0.
See all versions of nansen-cli for other recorded verdicts.
- Review the evidence and your use of nansen-cli@1.46.0 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.