Back to skill

Security audit

memory-system

Security checks for vulnerabilities and agentic risk

Overview

This memory skill is coherent and local-only, but it gives agents broad persistent write authority over memory and instruction files without clear approval boundaries.

Install only if you want an agent to keep local long-term memories. Review MEMORY.md and memory/ contents regularly, avoid storing secrets, and require explicit approval before it edits AGENTS.md, TOOLS.md, skill files, or any scheduled automation.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (8)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest lists several trigger phrases, but then adds the broad condition '当需要为新 Agent 配置记忆能力、或管理已有记忆时使用', which is open-ended and does not clearly define boundaries for when the skill should or should not activate. This can cause unintended invocation because '管理已有记忆' is a wide class of ordinary actions without exclusions or negative examples.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The file’s headings, comments, and operational instructions are entirely in Chinese, which imposes a specific language on users or maintainers. There is no indication that use of Chinese is optional, selectable, or required for a documented region-specific purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown template uses Chinese headings and instructions throughout, such as the title and section labels, which imposes a specific language on users. The file does not offer a language choice or explain that the template is intended only for a Chinese-language or region-specific context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill explicitly instructs the agent to create and update persistent local memory files that may contain user context, but it does not require user consent, visibility, or disclosure before writing. In an agent setting, silent persistence can capture sensitive personal or project information and retain it beyond the current session, creating privacy and data-handling risks even if the feature is intended for continuity.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instructions tell the agent to write learned lessons into AGENTS.md, TOOLS.md, or other skill files, which are configuration and instruction surfaces that can materially change future agent behavior. Without an explicit warning, approval step, or scope restriction, this creates a persistence channel for prompt/instruction poisoning and unauthorized modification of project control files.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The document directs the agent to modify persistent memory files and archive content without requiring an explicit user warning, confirmation, or approval boundary. In a memory-management skill, unattended writes can silently alter long-term state, causing data loss, unwanted retention, or integrity issues if the agent distills or archives incorrectly.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The automation section recommends scheduled automatic updates to persistent memory files, which increases risk because writes may occur without contemporaneous user awareness or validation. In this skill context, automated distillation can propagate summarization errors, overwrite important context, or archive needed information, making the absence of user warnings and approval controls security-relevant.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

A skill document that mandates a single language can create a locale policy issue when no user opt-in or justification is provided. This file presents all instructions exclusively in Chinese without stating that the skill is region-specific or offering an alternative language.

Content

No source excerpt is available for this finding.