Security audit

local-industrial-fund-analysis-writer

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent writing assistant, but its publication instructions conflict with its stated no-external-write boundary and could cause draft content to be sent to external document services without a clear confirmation gate.

Install only if you are comfortable using it as a draft-writing and public-source research assistant. Treat publication as a separate manual step: review the markdown first, remove sensitive or nonpublic information, and explicitly approve any Tencent Docs or other connector use before content leaves the workspace.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Description-Behavior Mismatch

Medium

Confidence: 87% confidence
Finding: The skill claims to be self-contained and not dependent on external experts or configuration, but elsewhere it directs use of external publishing tooling and web-based connectors. This mismatch can mislead operators and downstream policy systems about the skill’s actual trust boundary, causing unintended data flow to external systems or incorrect approval of the skill under a stricter isolation assumption.

Intent-Code Divergence

High

Confidence: 95% confidence
Finding: The skill states that publishing and external write actions are prohibited, but later prescribes publishing through another skill or connector. Contradictory action boundaries are dangerous because an agent may follow the more operational instruction and perform an external publication the user did not clearly authorize, leading to unintended disclosure of draft content or sensitive context.

Missing User Warnings

Low

Confidence: 80% confidence
Finding: The skill instructs use of web search on public sources without warning that user-provided topics, entities, or draft context may be sent to external services. In a research-writing workflow, prompts may contain unpublished investment interests, internal framing, or sensitive regional priorities, so silent transmission can create avoidable confidentiality and compliance risk.

Static analysis

No suspicious patterns detected.