Elite Longterm Memory
PassAudited by VirusTotal on May 12, 2026.
Overview
Type: OpenClaw Skill Name: elite-longterm-memory Version: 1.2.3 The skill bundle provides a multi-layered memory system for AI agents. While the `bin/elite-memory.js` script performs benign local file system operations, the `SKILL.md` contains instructions for the AI agent that involve executing external Python scripts (`python3 skills/git-notes-memory/memory.py`), installing npm packages (`npm install mem0ai`), and running external binaries (`supermemory`). These actions introduce supply chain risks and potential shell injection vulnerabilities if the external scripts or their arguments are not properly secured. Additionally, the `SKILL.md` instructs the agent to perform certain actions 'SILENTLY', which is a prompt injection technique, though in this context it appears intended for managing agent verbosity rather than hiding malicious activity. The `rm -rf` command is also present for a 'nuclear option' memory clear. These capabilities, while potentially useful for the stated purpose, carry significant risks without clear evidence of intentional malice.
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.
Old, incorrect, or sensitive memories may be reused later and shape the agent's behavior.
The skill intentionally creates persistent agent memory before replies, so stored context can influence future sessions.
Agent (internal): 1. Write to SESSION-STATE.md → "Decision: Use Tailwind" 2. THEN respond
Review and curate memory files regularly, avoid storing secrets, and remove or correct inaccurate memories.
The agent may update long-term memory without mentioning each update in chat.
The documentation tells the agent to store some memory silently, which reduces visibility into persistent changes.
# Store a decision (SILENT - never announce)
If you install it, consider instructing the agent to summarize memory changes or ask before storing sensitive information.
Private conversation or project context may leave the local machine if cloud memory integrations are enabled.
The Mem0 example shows conversation messages being sent to an external memory provider when that optional integration is configured.
await client.add(messages, { user_id: "user123" });Enable SuperMemory or Mem0 only if you trust the provider, understand retention policies, and avoid sending secrets.
A configured API key may incur costs or expose provider access if mishandled.
The skill requires a provider API key for its memory/search integrations, which is expected but grants access to the user's provider account.
Required env vars: OPENAI_API_KEY
Use scoped or dedicated API keys where possible, monitor usage, and revoke keys if you stop using the skill.
Running the command would delete stored vector memories and could cause loss of recalled context.
The documentation includes a destructive shell command that removes the LanceDB memory directory.
rm -rf ~/.openclaw/memory/lancedb/
Run this only as an intentional reset after backing up any memory data you may need.
Installing or running the package may also install third-party code.
The package can pull an optional external dependency with a version range; this is normal for npm integrations but still part of the supply chain.
"optionalDependencies": { "mem0ai": "^1.0.0" }Install from trusted sources, pin versions if needed, and review dependencies for sensitive environments.