Back to skill

Security audit

Dunning-Kruger Effect

Security checks for vulnerabilities and agentic risk

Overview

This coaching skill is proportionate to its purpose, with a privacy caution about saving examples from real use.

Keep saved examples general and anonymous, and require your approval before the agent adds interaction-derived material to the reusable skill. Avoid retaining personal, medical, employment, or confidential business details.

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 retain actual failure patterns from real interactions without warning users about persistent storage.

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 persisting interaction-derived information in the reusable skill without requiring user consent or sanitization. Because the skill addresses hiring, medical concerns, and personal performance, these entries could expose sensitive information to later users or repository readers. This is a conditional privacy risk: the file provides no automated storage mechanism, but an agent following the instruction with write access could create that exposure.

Content

No source excerpt is available for this finding.