Memory Vacuum

WarnAudited by ClawScan on May 10, 2026.

Overview

This memory-management skill is coherent, but it asks the agent to automatically persist and rewrite conversation memories without clear consent, retention, or rollback controls.

Review carefully before installing. This skill is not showing code execution or exfiltration, but it is designed to create persistent memory files from conversations and automatically rewrite them. Use it only if you are comfortable with broad chat logging, and prefer a version with clear opt-in saving, deletion controls, backups, and sensitive-data exclusions.

Findings (4)

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

Private or sensitive chat content could be stored persistently by default and later influence the agent, even if the user did not intend that message to become long-term memory.

Why it was flagged

The skill instructs the agent to append every message to persistent dated memory files, which can capture sensitive conversation content and reuse it across future sessions.

Skill content
| 写入日记 | 每条消息 | 追加到 `memory/YYYY-MM-DD.md` |
Recommendation

Only enable this if you want broad persistent memory logging. Add explicit opt-in, sensitive-data exclusions, a clear storage path, retention limits, and a user-review step before saving.

What this means

Incorrect, obsolete, or sensitive information may remain available to the agent indefinitely and affect later conversations.

Why it was flagged

The instruction says outdated memory is never deleted, creating indefinite retention of potentially stale, sensitive, or poisoned context.

Skill content
- 有过时信息 → 标记但不删除(永远不删除)
Recommendation

Provide deletion, expiration, and correction workflows, and let the user inspect or remove stored memory entries.

What this means

A bad summary or mistaken merge could permanently alter the memory used in future sessions, causing errors to propagate over time.

Why it was flagged

The skill tells the agent to automatically rewrite and compress shared memory before sessions, but does not describe backups, approval, rollback, or conflict handling.

Skill content
新会话开始前,检查 `MEMORY.md`:... 有重复信息 → 合并 ... 超过 200 行 → 压缩最旧的一半
Recommendation

Require user confirmation for bulk compression, keep backups or diffs, and make compression reversible.

What this means

Users may misunderstand whether their conversations are being persisted to disk.

Why it was flagged

The wording may be confusing: it says conversation history is not stored, while also saying raw conversations are written to daily diary files.

Skill content
- **不存对话历史**(太初靠上下文) ... - 原始对话 → 当日日记(按日期分流,不影响加载速度)
Recommendation

Clarify that the skill may store raw conversation logs in files, distinguish RAM use from disk storage, and explain privacy implications plainly.