use-outbridge

$npx mdskill add DiscreteTom/outbridge/use-outbridge

Operate local projects from a remote agent via Outbridge and SSH tunnel.

  • Enables remote agents to inspect, read, create, or modify local files.
  • Depends on the installed outbridge command and an SSH tunnel with VS Code port forward.
  • Reads live help and references setup.md only when configuring or repairing the connection.
  • Runs commands with local worker permissions, warning against unsafe retries after timeouts.

SKILL.md

.github/skills/use-outbridgeView on GitHub ↗
---
name: use-outbridge
description: Operate a local project from an agent running on a remote development machine through Outbridge and an SSH tunnel, including a VS Code Remote SSH port forward. Use when the remote agent needs to inspect, read, create, or modify local files, or run builds, tests, and shell commands on the local computer without requiring inbound local network access.
---

# Use Outbridge

Load this skill once before the first Outbridge task in a conversation or
session. Reuse these instructions for later Outbridge tasks instead of reading
the skill again.

## Workflow

Use the installed `outbridge` command. Read its live help and follow the
documented Agent workflow:

```sh
outbridge --help
```

Treat the top-level and subcommand help as the authoritative interface; read
the relevant help instead of guessing arguments. Read
[references/setup.md](references/setup.md) only when the user asks to configure
or repair the connection.

## Safety

- Treat the local computer and its files as the user's real environment.
- Do not overwrite files or run mutating commands unless the task requires it.
- Remember that Outbridge commands run with the local worker user's permissions;
  the workspace is not a sandbox.
- Do not retry a write or shell call automatically after a timeout or connection
  loss because its final state may be unknown.