T02 · Agent Memory Poisoning
Warning
- Location
SKILL.md:112- Finding
Unvalidated External Skill Patterns Can Poison Persistent Agent State
- Content
View full analysis
- Remediation
View remediation
Security audit
Shuxia Skill Library
Security checks for vulnerabilities and agentic risk
Overview
The skill is mostly coherent, but it tells agents to persist patterns from external skills into a hardcoded user-level Codex skill file without an approval or provenance step.
Review the persistence behavior before installing. This skill should be used only if you are comfortable with it updating a local skill-pattern library, and the destination should be changed to a user-approved path with preview, provenance, and approval before writes.
Vulnerability Patterns
- Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
- Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
- 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
Findings (1)
Vulnerability Patterns
- 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
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)
Missing User Warnings
Low
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
The skill explicitly instructs storing distilled ideas in a hardcoded local file path under the user's home directory. Even though this is documentation rather than executable code, it encourages file modification without any warning, confirmation step, or safeguard about overwriting user data, which can lead to unintended changes in the local skill library.
- Content