胡田 OPC导师 并购对接

• Outcome: Review
• Latest audit: May 28, 2026, 11:30 AM
• Version:
  1.0.0

Overview

This skill is mostly an M&A outreach playbook, but it includes guidance for hiding acquisition intent and using alternate roles to reach decision-makers.

Install only if you will use it with strict controls: disclose your real organization and commercial intent early, use only lawful public business contact channels, avoid personal or scraped contact data, honor opt-outs, limit follow-ups, and have legal or compliance review before using it for real M&A outreach.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (15)

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The skill explicitly recommends approaching targets under alternative identities such as '投资者/行业研究者身份' to open communication channels. That is deceptive pretexting rather than neutral business outreach, and it can facilitate social-engineering-style access to executives who would not otherwise engage on the true acquisition topic.

Context-Inappropriate Capability

Medium

• Confidence: 97% confidence
• Finding: The guidance to avoid mentioning M&A directly and to '绕道' around filtering by board secretaries is an access-evasion tactic. It encourages manipulation of gatekeepers to obtain meetings under incomplete or misleading pretenses, increasing the risk of deceptive business engagement and compliance issues.

Intent-Code Divergence

Medium

• Confidence: 84% confidence
• Finding: Although this section presents '合法合规渠道', the broader skill later undermines that framing by pairing lawful data sources with misleading outreach personas. This creates a contradictory workflow where legitimately sourced contact data can be used in deceptive approaches, materially increasing abuse potential.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill gives detailed instructions for identifying, collecting, validating, and recording personal and professional contact details of named individuals, but lacks a strong user-facing privacy and data-minimization framework. In practice this can encourage unnecessary collection, retention, and downstream use of personal data without clear lawful basis, notice, retention limits, or purpose restriction.

Missing User Warnings

Medium

• Confidence: 84% confidence
• Finding: This section provides a template and workflow for collecting, storing, and using individuals' contact details and intermediary relationship data, but it does not include concrete handling requirements such as consent checks, retention limits, access controls, or disposal rules. Although the document emphasizes public sources and lists some prohibited conduct, the skill context is M&A outreach, where personal and professional contact data can be aggregated and operationalized in ways that create privacy, compliance, and misuse risks.

Missing User Warnings

Medium

• Confidence: 84% confidence
• Finding: The document gives actionable outreach steps using public contact details and follow-up calls, but it does not pair them with concrete privacy, consent, anti-harassment, or jurisdiction-specific compliance guardrails. In an M&A targeting context, this can normalize repeated unsolicited contact and collection/use of personal or quasi-personal contact data without clear lawful-basis checks.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly guides users to identify key individuals, obtain contact details, and plan outreach, but omits privacy, consent, and lawful-source constraints. In an M&A context, this can encourage collection or misuse of personal/business contact information, potentially leading to privacy violations, unsolicited targeting, or non-compliant handling of personal data.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill explicitly includes helping users obtain contact information, but the quick-start guide does not warn about privacy, lawful sourcing, authorization, or prohibited collection of personal data. In a deal-sourcing and outreach context, this omission can normalize scraping, use of non-public personal contact details, or non-compliant outreach practices.

Ssd 4

Medium

• Confidence: 88% confidence
• Finding: The staged model emphasizes relationship cultivation, repeated follow-up, and steering targets toward progressively more sensitive business engagement, while elsewhere allowing non-obvious framing of intent. In context, this normalizes a playbook for socially engineering access to decision-makers rather than simply organizing legitimate M&A outreach.

Ssd 1

Medium

• Confidence: 97% confidence
• Finding: This section recommends using alternative personas to gain initial access before revealing the real acquisition-related purpose. That is classic pretext-based outreach and materially raises the risk that the skill is used for deceptive social engineering against corporate personnel.

Ssd 1

Medium

• Confidence: 97% confidence
• Finding: Encouraging users to frame outreach to board secretaries indirectly so acquisition proposals are not filtered is a deceptive bypass strategy. Because board secretaries are specifically tasked with screening sensitive communications, instructing users to evade that function increases the chance of manipulative and unauthorized access to senior decision-makers.

Ssd 4

Medium

• Confidence: 91% confidence
• Finding: The strategy principles promote role-matching, trust-first engagement, and gradual escalation toward harder-to-reach individuals in ways that can be used to mask the actor's true purpose. In a deal-origination skill, this is dangerous because it facilitates pretexting-style access and social engineering against gatekeepers and senior stakeholders.

Ssd 4

Medium

• Confidence: 95% confidence
• Finding: The cold-start guidance explicitly advises avoiding mention of 'M&A' and presenting as an industry researcher to secure initial engagement. That is a classic deceptive-access pattern: benign framing is used to bypass likely resistance to the real objective, increasing social-engineering risk.

Ssd 4

High

• Confidence: 98% confidence
• Finding: This section is especially risky because it explicitly recommends contacting a board secretary using an alternate identity and 'detouring' around filters instead of being direct about M&A intent. That is targeted guidance for bypassing organizational screening controls, which materially elevates the document from generic networking advice to operational social-engineering content.

Ssd 4

High

• Confidence: 99% confidence
• Finding: The risk-control section directly instructs users to conceal intent early and disclose it only after rapport is built. This is explicit stealth guidance that enables manipulation of the target's decision-making and makes the skill substantially more dangerous in the M&A-contacting context.

VirusTotal

VirusTotal findings are pending for this skill version.