glovebox

glovebox is a skill for Claude Code, Codex from segentic-lab/glovebox-mcp. It costs 277 tokens per session (2,064 once invoked), scanned A, original, MIT.

A skill for operating real desktop applications and web browsers inside isolated Glovebox windows.

In plain words
What is it for?
It is for filling out web forms, using browser-based tools, and controlling applications such as GIMP, Inkscape, or LibreOffice.
Why use it?
It lets an agent work with interfaces that require clicking, typing, screenshots, file uploads, or other visible actions.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one. Also seen: mentions AGENTS.md.

Good fit It is for filling out web forms, using browser-based tools, and controlling applications such as GIMP, Inkscape, or LibreOffice.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/segentic-lab/glovebox-mcp/glovebox
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add segentic-lab/glovebox-mcp --skill glovebox
Clone the repo
git clone --depth 1 https://github.com/segentic-lab/glovebox-mcp

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for glovebox

README.md
[![agentmods](https://agentmods.dev/badge/skills/segentic-lab/glovebox-mcp/glovebox/github.svg)](https://agentmods.dev/skills/segentic-lab/glovebox-mcp/glovebox)
Your own site
<a href="https://agentmods.dev/skills/segentic-lab/glovebox-mcp/glovebox"><img src="https://agentmods.dev/badge/skills/segentic-lab/glovebox-mcp/glovebox/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for glovebox

Your own site · 80×15
<a href="https://agentmods.dev/skills/segentic-lab/glovebox-mcp/glovebox"><img src="https://agentmods.dev/badge/skills/segentic-lab/glovebox-mcp/glovebox.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 277 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,064 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00277 $0.02064
Opus 5.5 $0.00111 $0.00826
Sonnet 5.5 $0.00055 $0.00413
Haiku 4.5 $0.00028 $0.00206

Measured yesterday against content hash 7ac241b9a85f, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

glovebox scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/glovebox/SKILL.md · 119 lines

How it starts

The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Driving apps with Glovebox

Glovebox gives you a real app in a real window that only you can see and touch: each instance is its own X display (:2, :3, …) with its own cursor, keyboard focus and clipboard, rendered as a window on the person's desktop. You act with pixels and keys, like a person would, and you look before you act. Every tool answers JSON with ok and, on failure, fix — the call that unblocks you. Read fix and do what it says before trying anything else.

First moves

  1. status() — version, the vision backend actually in use (vision.effective: local = OmniParser/ScreenParser on the GPU, basic = tesseract OCR, none = your own eyes), running instances, missing system deps, notes.
  2. list_instances() — what is already running (other agents' windows show launched_here: false and a title). Never drive another agent's instance; launch your own.
  3. launch_app("chromium") / launch_app("gimp", size="1600x1000") → use the returned instance id in EVERY call. Check warnings.
  4. wait_for_stable(instance) then screenshot(instance) — the app needs a moment to draw.

The loop

Look → act → verify. You cannot see the screen unless you take a screenshot or parse_screen; after every action that changes the screen, observe again. The cheap way: pass observe="screenshot" (or "parse") and a settle_ms on the action itself, or wait_for_stable / wait_for_text after navigation. Then confirm the result before the next step. Ids from parse_screen/find_elements go stale the moment the screen changes.

Grounding: how to know where to click

Vision mode How to find a target
local (GPU) find_elements("Save", instance, type="button") → click_element(id). Elements have real types (button/link/tab/input/checkbox/…) with the ScreenParser detector; parse_screen lists everything (capped at 300, truncated flagged).
basic (OCR) find_elements("Sign in", instance) finds text lines; icons without text are invisible to OCR — screenshot + click(x, y) for those.
none screenshot(instance), reason about pixel coordinates yourself, click(x, y). Zoom with region="x,y,w,h" for small text.

Read the full file on GitHub · 119 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday Changed · +11 lines 7ac241b9a85f
  2. 2d ago First seen · 108 lines · 277 tokens per session scan A ca88c4193b6a

Subscribe to this mod's changes

glovebox is a skill published in the GitHub repository segentic-lab/glovebox-mcp (13 stars, last pushed yesterday), licensed MIT. It adds 277 tokens to every session and 2,064 once invoked, about $0.0011 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-14.

Related

glovebox compared with skills from other repositories

glovebox next to 6 skills from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
gloveboxsegentic-lab/glovebox-mcpSkill13yesterdayMITA277Claude Code, Codex
trusty-squireTrusty-Squire/trusty-squireSkill3yesterdayMITA208Cursor
browserstickerdaniel/browserjackSkill56d agoMITA18Claude Code
signed-in-browsert3ratech/mcp-session-bridgeSkill23d agoApache-2.0A109Claude Code, Codex
post-to-hacker-newst3ratech/mcp-session-bridgeSkill23d agoApache-2.0A75Claude Code, Codex
melaya-setupmelaya-labs/melaya-mcpSkill12d agoApache-2.0A119Claude Code
ciso-phishing-mailbox-safetymelaya-labs/melaya-mcpSkill12d agoApache-2.0A88Claude Code