▍ Free cross-client agent skill · manifest checked

Does this repo need a persistent workspace?

Run a read-only evidence audit before choosing remote infrastructure. The plugin recommends a rebuildable remote-agent task, a persistent development shell, or more investigation.

No telemetry No scripts or hooks No Terminalbox upload Source visible before install
Install from GitHub

Use the universal installer or your client’s native plugin flow.

The public repository contains only manifests, this read-only skill, its documentation, and an MIT license.

  1. 1 Install with any compatible Agent Skills client
    npx skills add Anchras/terminalbox-tools --skill workspace-fit
  2. 2 Or install with GitHub Copilot CLI
    copilot plugin install Anchras/terminalbox-tools:plugins/remote-workspace-fit
  3. 3 Or add the marketplace in Claude Code
    claude plugin marketplace add Anchras/terminalbox-tools
  4. 4 Install in Claude Code
    claude plugin install remote-workspace-fit@terminalbox-tools
  5. 5 Ask your client to run the audit
    Use the workspace-fit skill to audit this repository without changing it.

    Claude native-plugin users can also run /remote-workspace-fit:workspace-fit.

Review the complete plugin source ↗ Review the indexed source and security scans ↗ Verify installation in GitHub’s docs ↗ Verify installation in Anthropic’s docs ↗
What it checks

Repository facts, not infrastructure folklore.

The skill reads orientation docs, tracked manifests, lockfiles, development scripts, container configuration, and CI files. It does not execute the workflow it finds.

01

Rebuildability

Can a clean clone restore tools and dependencies from committed instructions?

02

Live state

Do databases, watchers, queues, containers, or agent processes need to keep running?

03

Interaction

Is the valuable state an interactive shell or tmux session steered across devices?

04

External boundaries

Are private networks, SSH hops, non-GitHub sources, hardware, or local data required?

The decision boundary

The plugin is allowed to say “do not buy.”

Rebuildable remote task

A fresh clone restores the useful state.

Use your existing provider workflow when setup is repeatable and the task can return a branch or pull request asynchronously.

Persistent workspace

The useful state is still running.

Consider a persistent shell when processes, interactive state, or cross-device continuity are the actual requirement.

See the €6/month persistent option →
Evidence incomplete

The deciding facts are not in the repository.

Ask one focused question or run one safe validation step. The plugin does not recommend buying anything without repository-backed evidence.

Ask for an honest fit check →
Example shape

A short recommendation you can challenge.

Every result separates observed evidence from unknowns and ends with one safe next test.

Recommendation
Rebuildable remote task — the tracked lockfiles and CI setup restore the project.

Repository evidence
• package-lock.json and npm ci are committed
• CI provisions the only documented service
• no long-running local process is required

Persistence triggers
None found.

Next test
Confirm one fresh agent task can run the documented check.
Inspect before trust

A skill file, not an agent hiding a service.

Included

One Markdown skill with the audit workflow, decision rules, and output contract.

Not included

No executable scripts, hooks, MCP servers, agents, credentials, or network client.

Forbidden by the skill

Reading secret values, installing dependencies, starting services, or changing files.

Measurement

This page records bounded view and button labels. Plugin use and repository contents stay within your installed agent client and are not sent to Terminalbox.