ClawHub
Back to skill

Security audit

Security Equity Research

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only equity research template skill with overstated automation wording, but no executable install code, tool access, persistence, or hidden data handling.

Treat this as a template and formula reference, not a live market-data connector. Do not provide nonpublic financial data, customer information, personal account details, or insider material; verify all market, regulatory, and valuation assumptions independently and route any publishable research through qualified analyst and compliance review.

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
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill presents a hard boundary claiming no data access or tool use, but later advertises 'automatic retrieval' from financial, macro, and industry databases. That contradiction can cause unsafe routing and user over-trust: operators may expose sensitive or licensed data under the assumption the skill can safely and compliantly ingest it, or downstream systems may later wire in retrieval behavior inconsistent with the stated controls.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The execution-boundary section explicitly states that dcf_valuation() and similar blocks are only illustrative, that the skill 'does not run' and 'does not fetch data'. However, later code blocks define working Python functions and include a usage example that invokes dcf_valuation(...) and prints computed outputs, which directly contradicts the stated non-execution framing in the documentation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill presents a hard boundary claiming no data access or tool use, but later advertises 'automatic retrieval' from financial, macro, and industry databases. That contradiction can cause unsafe routing and user over-trust: operators may expose sensitive or licensed data under the assumption the skill can safely and compliantly ingest it, or downstream systems may later wire in retrieval behavior inconsistent with the stated controls.

Content

No source excerpt is available for this finding.