RMN Visualizer

Visualize your AI Agent's 5-layer recursive memory network from local files with an interactive D3.js force-directed graph in a Node.js server.

This skill will read markdown files from your OpenClaw workspace (MEMORY.md, memory/*.md, .issues/*, SOUL.md) and can start a Cloudflare tunnel that publishes a public URL serving those parsed files. Before running or letting an agent auto-run the 'launch' flow: 1) Inspect the workspace files to ensure no secrets or private data are present; 2) Prefer running node scripts/serve.js locally and opening http://localhost:3459 rather than using the tunnel; 3) If you must use the tunnel, run it yourself manually and do not instruct the agent to automatically post the URL to chat; 4) Consider copying a sanitized subset of memory files into a temporary test workspace to preview behavior; 5) If you do not trust automatic sharing, do not grant the agent permission to invoke this skill autonomously.

Type: OpenClaw Skill Name: rmn-visualizer Version: 1.1.2 The skill is classified as suspicious due to its core functionality involving the reading of potentially sensitive agent memory files (MEMORY.md, memory/*.md, .issues/*, SOUL.md) and then exposing this processed data to the public internet via a Cloudflare Tunnel, as orchestrated by `scripts/launch.js` and `scripts/serve.js`. While this aligns with the stated purpose of visualizing agent memory, it creates a significant vulnerability risk if the agent's memory contains confidential information. The `SKILL.md` also contains a functional prompt injection instruction to the agent to output only the public URL, which, while intended for user convenience, is a signal of agent instruction manipulation.