Security audit

Memory Management

Security checks for vulnerabilities and agentic risk

Overview

This skill persistently manages project memory, but its sensitive actions are disclosed, local, and gated by user authorization.

Before installing, understand that this skill is meant to create and maintain persistent local project memory. Use it only in repositories where storing project context is acceptable, confirm paths before writes or purges, and be especially cautious with cross-project comparisons or memory containing personal, credential, or production data.

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

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The manifest describes this skill as managing HOT/WARM/COLD working memory, reviewing saved findings, reconciling notes, and handling authorized erasure while explicitly avoiding canonical registry changes. Example 1 instead presents the skill as updating keyword rankings, optionally running an automated ranking check, and generating SEO recommendations, which is a broader operational/analytics function rather than memory management itself.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The examples show persistent file updates and creation of ranking history without warning the user that durable memory is being modified. In a memory-management skill, silent persistence is especially risky because users may reveal business-sensitive data assuming the interaction is ephemeral, leading to privacy, retention, and governance issues.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The glossary lookup trigger is phrased like ordinary conversation and lacks an explicit activation boundary, making accidental invocation plausible during normal chat. In a memory-management skill, this can expose stored project context or cause the agent to retrieve organizational data without a clearly scoped request.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 88% confidence
Finding: The initialization example can be triggered by a broad natural-language request and leads to creation of persistent memory structures. Without clear activation constraints and confirmation, a user could unintentionally cause storage initialization, creating durable records and project scaffolding that outlast the conversation.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Low

Category: Not specified by scanner
Confidence: 76% confidence
Finding: The manifest focuses on authorized working-memory lifecycle actions for a subject or project context, with privacy and registry-ownership constraints. The documented features add broader behaviors such as cross-project memory comparison and generalized memory health checks, which extend beyond the manifest's explicitly described user intents and governance boundaries.
Content: No source excerpt is available for this finding.

Vague Triggers

Low

Category: Not specified by scanner
Confidence: 86% confidence
Finding: The advanced feature commands are imperative and broad, but they do not define who may invoke them, what project scope applies, or what safeguards exist for sensitive operations like cross-project comparison and bulk state changes. In this skill context, ambiguous triggerability increases the chance of unauthorized data movement, over-collection, or unintended memory mutations.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.