Security audit

Lobster

Security checks across malware telemetry and agentic risk

Overview

Lobster is a disclosed workflow runner with powerful shell, state, and Clawdbot-invocation features that are risky if misused but coherent with its stated purpose.

Install only if you trust the external Lobster CLI or npm package source. Treat Lobster pipelines like scripts: review shell commands, workflow files, and clawd.invoke calls before running them; use scoped CLAWD tokens; add approval gates before side effects; and relocate or clear ~/.lobster/state/ if stored workflow state may be sensitive.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Behavioral ASTexec() Call, eval() Call, Dynamic Import
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The documentation presents shell execution and persistent state operations as normal commands without prominently warning that they can execute arbitrary system commands and write durable local state. In an agent-skill context, that omission can cause unsafe use, especially if an operator assumes the runtime is intrinsically safe because it supports approval gates.

Missing User Warnings

Low

Confidence: 80% confidence
Finding: The documentation shows credential-backed external tool invocation using `CLAWD_URL` and `CLAWD_TOKEN` but does not warn about token sensitivity, outbound requests, or the possibility of side effects such as sending messages. This is dangerous because users or agents may invoke it without understanding that it can act on remote systems using privileged credentials.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.