每日记忆管家

• Outcome: Review
• Latest audit: May 29, 2026 · 9:24 PM UTC
• Version: 1.0.2

Overview

This memory skill is not deceptive or destructive, but it is designed to automatically save recent session content into persistent local files without enough consent, retention, or sensitive-data controls.

Install only if you deliberately want recent OpenClaw session activity summarized into local daily memory files. Prefer manual use or add explicit confirmation, secret/PII redaction, retention limits, and a clear way to inspect, disable, and delete ~/.openclaw/workspace/memory entries before enabling cron mode.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 95% confidence
• Finding: The skill describes and embeds behavior that reads and writes local files and uses shell commands, but it declares no permissions. That creates a hidden capability surface: operators and users cannot accurately assess what the skill can access, and the skill can persist data or invoke commands without explicit approval boundaries. In this context, the undeclared write access is especially risky because it stores session-derived content to disk automatically.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 89% confidence
• Finding: The documented purpose understates the actual behavior by omitting important side effects such as local file creation, config handling, and external session enumeration. This mismatch can mislead users into enabling the skill without understanding that it will create files and inspect recent session data, reducing informed consent and review effectiveness. The risk is amplified because the skill runs automatically on a cron schedule.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: Manual trigger phrases like "remember this" and "update memory" are common conversational language and may be invoked unintentionally during normal interaction. That can cause the skill to persist content the user did not intend to save, including sensitive snippets from the surrounding conversation. Because the destination is a long-term memory file, accidental activation has durable privacy consequences.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill automatically records session content to persistent daily files, but the description does not prominently warn users that their conversation data may be stored long-term. Without clear notice, users may share sensitive information under the assumption that the interaction is ephemeral. In a memory-keeping skill, lack of explicit disclosure materially increases privacy and compliance risk.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The description advertises a manual trigger phrase, 'remember this', that is generic enough to appear in ordinary conversation. In an agent environment, broad trigger phrases can cause unintended activation, leading the skill to capture or persist session content without deliberate user intent. The memory-logging context makes this more dangerous because accidental invocation may store sensitive or private data into long-term memory files.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: Automatic summarization of recent sessions into persistent memory files can capture credentials, personal data, proprietary information, or other sensitive user-provided content without robust safeguards. The skill includes only lightweight noise filtering and deduplication, which do not meaningfully prevent sensitive-data retention. Since it runs hourly and writes long-term records, this creates an ongoing privacy and data-governance exposure.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.