tavily-remote-mcp

Fail

Audited by Socket on Mar 10, 2026

3 alerts found:

Obfuscated Filex3
Obfuscated FileHIGH
SKILL.md

The skill presents a coherent purpose—access as a unified OneKey Router to perform web-searching, extraction, crawling, mapping, and research. Its footprint is proportionate to that purpose: it requires an API key, does not request broad cloud credentials, and operates via documented tool scripts. Primary concerns center on credential handling (the OneKey access key) and data flows (outbound searches and content processing) which could expose the key or collected data if not properly isolated, logged, or redacted. The fallback demo key is a notable anomaly that should be restricted to testing environments. Overall, the risk is cautious to moderate (securityRisk around 0.40) with elevated attention to credential exposure and data flow integrity.

Confidence: 98%
Obfuscated FileHIGH
scripts/tavily_extract.py

The file itself does not contain traditional malware constructs (no reverse shell, no dynamic code execution, no obfuscation). However, it forwards unvalidated user-supplied or file-read JSON to an external agent via OneKeyAgentRouter.invoke and embeds a hardcoded default access key. Primary risk is inadvertent data exfiltration of sensitive local data and misuse of the embedded key. Treat the external library call as a high-risk sink until the third-party implementation and endpoints are audited; remove the default credential and add stricter input validation before deploying.

Confidence: 98%
Obfuscated FileHIGH
scripts/tavily_map.py

The script itself is not directly malicious: it performs JSON input parsing, minimal validation for a 'url' field, constructs a router using an environment-provided or fallback key, and forwards the payload to a remote OneKeyAgentRouter.invoke call. Primary risks are supply-chain and data-exfiltration: (1) a hardcoded fallback API key present in source (security smell and potential leakage/backdoor if active), and (2) full trust in the ai_agent_marketplace dependency which performs the network communication. Treat this module as a moderate security risk until the external dependency and the remote endpoint behavior are audited; do not send sensitive credentials or secrets through this tool and remove the hardcoded key.

Confidence: 98%
Audit Metadata
Analyzed At
Mar 10, 2026, 08:47 AM
Package URL
pkg:socket/skills-sh/aiagenta2z%2Fonekey-agent-router%2Ftavily-remote-mcp%2F@dfce776a54526daa58615aaf55a637e58ab6db8b
Security Audit — socket — tavily-remote-mcp