ClawHub
Back to skill

Security audit

投标文件编制助手

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed bid-document formatting helper that locally parses and writes Word files, with optional knowledge-base checks, and I found no hidden, destructive, or unrelated behavior.

Before installing, expect this skill to process uploaded .docx tender files and create or modify local .docx/config outputs. Use explicit output filenames, keep backups before running the repair script in overwrite mode, and only enable knowledge-base checks when you are comfortable with the platform using tender formatting details for lookup.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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
88% confidence
Finding
The skill documents local code execution that writes files (generation of .docx and XML repair) but does not declare the corresponding permission boundary. Undeclared write capability weakens least-privilege controls and can let a host agent invoke filesystem-affecting behavior without explicit policy review.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The skill claims it is limited to local document skeleton generation, but later introduces optional knowledge-base retrieval. This mismatch can cause an orchestrator or user to trust the skill as offline/local-only when it may access external or platform retrieval sources, creating data-flow ambiguity and possible unintended disclosure of tender contents or metadata.

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The hard-boundary section states the skill will not autonomously initiate external API or tool calls, yet a later section authorizes knowledge-base retrieval under certain conditions. Contradictory safety rules are dangerous because downstream agents may follow the more permissive instruction path, bypassing expectations about isolation and causing unreviewed external access.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger phrases are broad enough to match general bid-document creation requests, while the skill is only intended to generate formatting skeletons. This can cause the skill to activate outside its safe scope, potentially steering users toward tender-document drafting workflows that may implicate compliance-sensitive or substantive bid content even if the manifest says content is format-only.

VirusTotal

47/47 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.