T02 · Agent Memory Poisoning
Error
- Location
SKILL.md:7- Finding
Untrusted Content Can Be Stored as Permanent Agent Rules and Persona Memory
- Content
View full analysis
- Remediation
View remediation
Security audit
总结
Security checks for vulnerabilities and agentic risk
Overview
The skill is a disclosed memory-summarization workflow, but it gives agents broad persistent memory-writing guidance without enough safeguards for sensitive content or prompt-like instructions.
Review proposed memory entries before saving, especially P0/L0 or shared entries. Do not store raw logs, tokens, passwords, private URLs, customer data, or untrusted text that looks like instructions; sanitize summaries and keep external-source material separate from trusted rules.
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
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- 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 (2)
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
The skill explicitly instructs persisting user/session-derived content into a long-lived memory system and named files, but provides no guidance to detect, minimize, or redact sensitive data before storage. Because summaries of incidents, preferences, troubleshooting, and decision records can easily contain personal data, credentials, internal URLs, or business-sensitive context, this creates a real risk of unnecessary retention and later disclosure.
- Content
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The skill specifically tells the user to record error logs, commands, and configuration details, which are common sources of secrets such as API tokens, usernames, hostnames, file paths, cookies, and embedded credentials. Persisting these artifacts to memory files without warnings or redaction steps materially increases the chance of credential leakage, privacy violations, and exposure of internal infrastructure details.
- Content
Static analysis
No suspicious patterns detected.