Memory Transfer Enhanced

ReviewAudited by ClawScan on May 10, 2026.

Overview

The skill mostly matches its memory-transfer purpose, but it can copy persistent memories, including personal information, between agents and has weak path and privacy boundaries.

Only install this if you intend to move memory between agents and are comfortable reviewing the transferred content. Use interactive or dry-run workflows, prefer narrow topic/file transfers, avoid clone mode for private data, and verify source/target agent IDs carefully.

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

A crafted or mistaken agent ID could make the script read from or write under locations outside the intended OpenClaw workspace area, subject to local file permissions.

Why it was flagged

The source and target agent IDs are command arguments, and the fallback workspace path joins the raw agentId to /home/node/.openclaw without visible validation or a path containment check.

Skill content
return path.join(WORKSPACE_BASE, agentId);
Recommendation

Validate agent IDs, reject absolute paths and '..' segments, and enforce that resolved source and target paths stay under approved /home/node/.openclaw/workspace-* directories.

What this means

Private details or harmful instructions stored in one agent's memory may be carried into another agent and reused in future conversations or tasks.

Why it was flagged

The skill intentionally copies persistent memory content between agents, including a mode that preserves user information and any embedded instructions verbatim.

Skill content
Memory Cloning (clone mode) Copies memory exactly as-is without filtering or transformation: - All content preserved verbatim - User information remains
Recommendation

Use topic/file filters and dry-run first, manually inspect memory content before transfer, avoid clone mode for sensitive memories, and back up the target agent's memory before writing.

What this means

Users may believe share mode fully removes personal information when sensitive details can still be copied to another agent's persistent memory.

Why it was flagged

The documented share mode says it filters user information, but the visible implementation relies on a limited set of regex patterns that will not reliably remove all names, emails, phone numbers, addresses, or preferences.

Skill content
const USER_INFO_PATTERNS = [ /我的名字叫[^\n,。,.]+/g, /我叫[^\n,。,.]+/g, /username.*?:.*/gi, /name.*?:.*/gi, /我的邮箱是[^\n,。,.]+/g, /我的电话[^\n,。,.]+/g, /我的地址[^\n,。,.]+/g ]
Recommendation

Describe filtering as best-effort, show a content preview before writing, and require explicit user approval after reviewing the transformed output.

What this means

It is harder to confirm that the reviewed files correspond exactly to the advertised registry package and version.

Why it was flagged

The packaged metadata does not match the registry-provided slug/version for 'memory-transfer-enhanced' version 2.1.0, creating provenance and review-context ambiguity.

Skill content
"slug": "memory-transfer", "version": "1.0.0"
Recommendation

Verify the package source and metadata before installing, and prefer a release whose packaged metadata matches the registry listing.