Security audit

Clawsec Monitor

Security checks across malware telemetry and agentic risk

Overview

ClawSec Monitor is a disclosed local traffic-inspection proxy with serious privacy implications, but the sensitive behavior matches its stated purpose and is not hidden.

Install only if you intentionally want a local proxy that can inspect AI-agent traffic. Prefer per-process proxy and CA trust settings, avoid system-wide CA trust on production machines, restrict Docker exposure to localhost, and delete /tmp/clawsec plus remove any trusted CA when finished.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The code actively performs HTTPS man-in-the-middle interception by generating a CA and signing per-host certificates, enabling decryption and inspection of plaintext traffic. In an agent skill context, this is highly sensitive because it can capture prompts, tokens, API keys, session cookies, and other confidential data from all proxied HTTPS connections.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The README explicitly promotes full HTTPS interception and decrypted traffic inspection, which creates significant privacy and data-handling risk because the proxy can observe credentials, personal data, tokens, and other sensitive contents in transit. Although the document later includes limited cautions, the warning near the interception instructions is not sufficiently prominent or explicit about the scope of captured sensitive data and the operational/legal implications of doing MITM on agent traffic.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The skill prominently advertises interception and logging of all AI agent traffic, including snippets that may contain secrets, but does not present a strong up-front privacy warning or consent guidance before use. This can lead operators to deploy traffic capture without adequately understanding that sensitive prompts, credentials, and file contents may be recorded to disk.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill documents installation of a locally generated CA into the system trust store, which materially expands the proxy's ability to intercept HTTPS across trusted applications if misused or left installed. Although it says 'review carefully,' it lacks a prominent warning explaining the security consequences of trusting a local root CA and the need to remove it after use.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: README.md:112

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: run_tests.py:345

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: skill.md:157