ClawHub
Back to skill

Security audit

Advanced Human Writing & AI Humanizer

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed local writing and audit toolkit with some file-writing and optional MCP/plugin behavior that users should configure carefully.

Before installing, treat this as a broad local writing toolkit: use trusted project folders, review fix previews before using --apply, choose a trusted Python interpreter for the DeepSeek Harness plugin, and do not expose the MCP HTTP server beyond loopback unless you provide proper authentication.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding
The skill advertises executable components, CLI commands, MCP coordination, and file-processing workflows, but the manifest does not declare permissions or capability boundaries. That mismatch can cause the host or user to invoke a skill with broader file, shell, network, environment, or MCP access than expected, increasing the chance of unauthorized data access, prompt injection propagation through external files, or dangerous command execution.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The plugin allows the executable used to launch the MCP server to be fully controlled by config.pythonCommand or the PYTHON environment variable. That means a caller can cause this writing skill to run an arbitrary program under the guise of starting a Python backend, which expands the skill from text processing into general command execution. In the context of a writing/audit skill, this is broader than necessary and increases risk if configuration is user-influenced or untrusted.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The manifest description contains an extremely broad trigger list spanning many common writing and editing requests, which can cause the skill to be auto-selected in contexts far beyond its safest intended use. Because this skill also references executable and multi-agent workflows, overbroad activation raises the likelihood that sensitive user content is routed into unnecessary tooling or broader file-processing pipelines.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The code writes to either the original source file (`apply=True`) or an arbitrary `output_path` without any validation, backup, or confirmation step. In an agent/tooling context, this can unintentionally overwrite important files if a caller supplies a sensitive path or enables in-place editing on the wrong file, leading to data loss or corruption.

VirusTotal

VirusTotal engine telemetry is currently stale for this artifact.

View on VirusTotal

Static analysis

Detected: suspicious.dynamic_code_execution

Dynamic code execution detected.

Critical
Code
suspicious.dynamic_code_execution
Location
tests/test_skillhub_package.py:12