Workspace Explorer

Securely share your workspace with your owner via a remote VS Code environment. Use when (1) the owner requests to view or inspect your working files, (2) you need to give the owner live access to browse your codebase, (3) the owner wants to install extensions or use IDE features to explore files, or (4) you need a temporary secure tunnel for remote workspace inspection.

This skill does what it says (creates a public browser VS Code to let someone inspect files), but that capability inherently lets a remote person see or edit anything in the served directory — including secrets. Before using: (1) inspect the GitHub repo yourself (start_workspace.py and any scripts it downloads) and prefer running it in an isolated environment (ephemeral VM or container); (2) only serve a directory that contains no credentials, tokens, SSH keys, or other sensitive files; (3) verify the sources and checksums of downloaded binaries (code-server, cloudflared) or install them from official package channels yourself instead of letting the script fetch them; (4) share the URL/password only over a trusted channel and terminate the session when done; (5) monitor logs/cloudflared.log and remove downloaded binaries when finished. If you cannot audit the remote script or binaries, treat this skill as risky and avoid running it on any machine containing sensitive data.

Type: OpenClaw Skill Name: workspace-explorer Version: 1.1.5 This skill is suspicious due to its high-risk capabilities, despite aligning with its stated purpose. It instructs the agent to download and execute external binaries (code-server and cloudflared) from an external GitHub repository (https://github.com/mrbeandev/workspace-explorer.git) as described in SKILL.md, which introduces a significant supply chain risk. Furthermore, it creates a public internet tunnel via Cloudflare to expose the agent's workspace, providing remote interactive access, which is an inherently high-risk operation. While the prompt instructions in SKILL.md and HEARTBEAT.md are aligned with the skill's function (sharing the URL/password, checking status), the underlying actions carry substantial security implications.