ClawHub
Back to skill

Security audit

AI KEY·从哪儿下手

Security checks for vulnerabilities and agentic risk

Overview

This goal-planning skill needs review because it saves personal and business observations for future use without clear user control.

Before use, review the missing configuration and shared instructions. Require approval of each saved note, exclude confidential details, and establish how to inspect and delete memory and reference-library additions. The supplied static scan reported no findings, but the inspected instructions support the retention concern.

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
  • 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 (2)

The skill instructs the agent to persist user information without requiring user-facing disclosure.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The memory instructions require saving user-specific observations to a configured persistent directory without requiring disclosure or checking whether the user authorized retention. In this business-planning context, those observations can reveal financial circumstances, recurring difficulties, or confidential business mechanisms; local storage limits the exposure compared with external transmission, but still creates an undisclosed retention risk.

Content

No source excerpt is available for this finding.

The workflow instructs the agent to persist user-related information in memory without requiring a user-facing privacy notice.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

Question 29 directs the agent to write learned information into memory and marks that step as non-skippable, without requiring a user-facing notice or authorization check. Labeling entries as user statements, platform data, or AI inference improves provenance but does not address whether potentially sensitive user or business information should be retained.

Content

No source excerpt is available for this finding.