Skip to content
Back to skills

Foxglove

ASecurity

Visualize live or recorded ROS robot data remotely with Foxglove or Lichtblick.

  • 34 stars
  • 0 votes
  • 0 copies
  • 6 views
  • Added September 19, 2026
ai-agentsgodebugginggitdocumentation

Works with

  • cli

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned September 19, 2026

npx -y skills add robium-ai/robium --skill foxglove --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Foxglove?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Foxglove
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/robium-ai-foxglove/badge)](https://www.skillsdirectory.com/skills/robium-ai-foxglove)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: foxglove
description: Visualize live or recorded ROS robot data remotely with Foxglove or Lichtblick.
---

# Foxglove and Lichtblick

Treat remote visualization as a chain: ROS graph, bridge, network, client, and
layout. Find the first link without evidence.

## Choose the client deliberately

- **Lichtblick:** prefer an inspectable open-core viewer, local/offline use, or
  a browser client you can host and extend yourself.
- **Foxglove:** prefer its managed data, fleet, organization, and support
  services, or when the team already uses its layouts and platform.
- Both clients can consume a Foxglove WebSocket and recordings for the common
  ROS workflow. Re-check current licensing and self-hosting terms before making
  them an architecture dependency.

Read [client choices](CLIENTS.md) when licensing, hosting, extensions, or
managed services affect the decision. If the visualization tool itself has not
been chosen, use `visualization` first.

## Trace the live-data chain

- Prove the ROS topics and QoS locally before blaming the bridge.
- Run the bridge near the ROS graph; DDS discovery should not need to cross a
  browser, NAT boundary, or public internet.
- Keep its unauthenticated WebSocket on a private network, VPN, or SSH tunnel.
- Confirm the listening port and WebSocket reachability before debugging a
  client layout.
- Limit high-rate topics at the bridge when bandwidth is constrained. Hiding a
  panel does not reduce transport load.
- Save a focused layout per task and record before any run that must be shared,
  replayed, or compared.

For bridge, MCAP, navigation goals, teleop, actions, layouts, and custom clients,
read [ROS 2 workflows](ROS2-WORKFLOWS.md). For a connection that is blank,
unreachable, mixed-content blocked, or missing controls, read
[failures](FAILURES.md).

Cross into `ros2` when the local graph or QoS is wrong. Cross into
`environments` when the bridge host, VPN, or container cannot provide the
network contract. Cross into `navigation` only after pose, map, sensors, and
goal messages are proven across the viewer boundary.

## Done

- The intended client reaches the bridge without exposing it publicly.
- The necessary topics render at useful rates and with correct frames.
- Controls publish to the interfaces the robot actually consumes.
- The layout and recording path can be reused by another person.
- Current behavior and hosting terms were checked against official
  [Foxglove](https://docs.foxglove.dev/docs) and
  [Lichtblick](https://lichtblick-suite.github.io/docs/) documentation.

Files in this skill

  • CLIENTS.md2 KB
  • FAILURES.md2.2 KB
  • ROS2-WORKFLOWS.md2.7 KB
  • SKILL.md2.5 KB
  • evals.yaml437 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.