Back to skill

Security audit

CFO全球超级专家系统

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed CFO sub-skill router with high-impact finance/legal subject matter, but the inspected package does not show hidden execution, persistence, exfiltration, or destructive behavior.

Install only if you want a CFO-style router that may load other installed finance, legal, audit, strategy, and stock-analysis skills. Treat outputs as research support, not professional advice, and review any downstream sub-skill before allowing scripts, realtime market-data calls, or jurisdiction-specific compliance guidance.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The description markets a full-featured CFO super-expert system with multiple domain-specific engines and expert methodologies. However, the supplied code only defines a static registry of skill metadata, scenario groupings, and simple CLI commands to list, search, inspect, and count those skills. The only filesystem interaction is checking for the presence of SKILL.md files in subdirectories. There is no evidence in this code chunk of performing financial analysis, investment decisions, valuation, report/PPT generation, compliance reasoning, or orchestration beyond recommending related scenarios. This is a material description-versus-behavior mismatch: the code is a router/index for sub-skills, not the expert system itself.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
76% confidence
Finding

The skill instructs the agent to read sub-skill files from /data/user/skills/<sub-skill>/SKILL.md and to invoke local scripts, but it declares no explicit tool scope or permissions. This creates an authority/expectation gap where the runtime may permit filesystem or environment-backed operations without the skill clearly constraining what it can access, increasing the risk of unintended data exposure or overbroad execution if routing inputs are manipulated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The routing keywords are broad business terms like 财务管理, 审计, 合规, 战略, 并购, and 路演, which can match ordinary enterprise queries and trigger this skill or downstream sub-skills unexpectedly. In a multi-skill agent, overbroad activation can expand the effective capability surface, causing the system to load unrelated prompts, access additional local skill files, or run scripts when the user did not intend that path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The instruction states that all operational advice must comply with Chinese laws and regulatory requirements, but the skill is presented as a 'global' CFO expert system. This creates a natural-language locale policy issue because users are not offered a jurisdiction choice and the constraint is not framed as a clearly region-specific tool limitation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The registry embeds mixed English/Chinese keywords throughout and includes descriptions such as 'bilingual' in skill metadata, but there is no natural-language indication that language choice is optional or user-selectable. This can create an implicit locale preference in routing behavior rather than offering a documented opt-in.

Content

No source excerpt is available for this finding.