Back to skill

Security audit

compound-interest

Security checks for vulnerabilities and agentic risk

Overview

The skill provides compound-interest coaching but also directs the agent to save observations from every use without requiring your approval.

Before installing, remove the instruction to record observations after every use or require approval for each addition. Keep any approved notes generalized and free of personal or financial details. The coaching workflow does not require modifying the installed skill.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (1)

The skill instructs agents to persist real-use failure patterns without warning users about recording interaction-derived information.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The instruction to add observed entries after each real use and 'paste the actual failure pattern' encourages recording interaction-derived information in a reusable skill document without consent or sanitization requirements. In a financial-planning context, those entries could contain sensitive financial or personal details that later become accessible to other users or enter shared repositories. Exploitation depends on the agent having write access and following this instruction; the file does not explicitly direct external transmission.

Content

No source excerpt is available for this finding.