ClawHub
Back to skill

Security audit

招标文件版本智能对投标人版Bidding Doc Version Smart Compare Bidder

Security checks across malware telemetry and agentic risk

Overview

This skill is a local tender-document comparison assistant with some scope hygiene issues, but no hidden exfiltration, persistence, or destructive behavior was found.

Install only if you intend to analyze tender/bidding documents and are comfortable with uploaded document text being parsed locally and summarized with IMA knowledge-base lookups. Treat the output as an initial screening report, not legal advice, and be aware that the package contains a stale tenderer-report option and a lenient text-file fallback outside the documented DOCX/PDF-only path.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
86% confidence
Finding
The file explicitly documents and supports a tenderer-facing mode even though the skill metadata says it is only for bidder-side analysis. This creates a scope/behavior mismatch that can expose unintended functionality, confuse downstream policy enforcement, and produce outputs for a user role the skill was not approved to serve.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The CLI and rendering path allow callers to select '--role tenderer', making the out-of-scope mode operational rather than merely documented. In an agent setting, this can bypass product-level expectations or authorization assumptions and generate reports for an unintended audience with different legal/compliance implications.

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The implementation accepts .txt, .md, .markdown, and even arbitrary unknown extensions by falling back to plain-text extraction, which exceeds the skill’s declared DOCX/PDF-only scope. In this skill context, scope expansion is security-relevant because downstream components may assume only vetted document types are processed; this can enable prompt/data injection through unexpected text formats, bypass policy or product restrictions, and weaken user trust about what inputs are handled.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The trigger example phrase is broad enough to overlap with ordinary user requests about comparing two documents, which can cause the skill to activate in situations beyond its narrow intended scope. While this is not a code-execution or data-exfiltration issue, over-triggering can route unrelated documents into a bidding-analysis workflow, increasing the chance of confusion, incorrect analysis context, or unnecessary exposure of sensitive files.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.