Remember

PassAudited by VirusTotal on May 12, 2026.

Overview

Type: OpenClaw Skill Name: remember Version: 1.0.0 The skill bundle is designed for AI agent memory management, which is a benign purpose. However, the `categories.md` file suggests the agent should remember 'keys locations' within an `environments.md` file. While this instruction does not explicitly command the agent to read or exfiltrate sensitive data, it creates a risky capability by instructing the agent to store pointers to potentially sensitive information, making it a higher-value target for subsequent prompt injection attacks. This constitutes a potential vulnerability rather than direct malicious intent.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

The agent may remember preferences, corrections, commitments, relationships, or project context and reuse them later, which can be helpful but may also preserve outdated or private information.

Why it was flagged

The skill directs the agent to turn conversation events into persistent memory entries. That is the skill's stated purpose and includes metadata guidance, but persistent memories can influence future sessions and may contain sensitive context.

Skill content
During Session
Flag important moments with metadata:
- User corrects you  corrections.md + date + confidence:certain
- User states preference  preferences.md + source:explicit
Recommendation

Before using it, decide what categories are acceptable to store, use the documented 'forget' and 'never remember' controls, and avoid saving secrets, health details, or sensitive work information unless you intentionally want them retained.