glass AGENTS.md

Repository instructions for glass, a Rust server that lets an AI agent launch, view, control, and debug native desktop and mobile apps through MCP, a standard way for software tools to communicate.

In plain words
What is it for?
Use them when changing glass code, choosing its platform backends, running builds or tests, handling app screenshots and logs, or preparing commits.
Why use it?
They keep work consistent with the project's public/private boundary, supported platforms, commands, layout, and Git workflow.

Instructions file for CodexOpenCode

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.

agentmods
npx agentmods add instructions/fixed-width/glass/agents-md
Clone the repo
git clone --depth 1 https://github.com/fixed-width/glass

Made for: Codex, OpenCode.

Per session 1,112 This file is loaded in full into every session.
When invoked 1,112 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.01112 $0.01112
Opus 5 $0.00556 $0.00556
Sonnet 5 $0.00222 $0.00222
Haiku 4.5 $0.00111 $0.00111

Measured yesterday against content hash 5869837398a9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

glass AGENTS.md 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.

Origin

Copies of this mod

1 near-identical copy found in the catalogue:

AGENTS.md · 78 lines

How it starts

The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

Guidance for AI Agents working in the glass open-core Rust workspace.

This repository is PUBLIC. Keep internal business/strategy, licensing-strategy framing, planning and design specs, and the internal security threat model out of this repo — that material lives in the private fixed-width repo. Document what the product does, never why it helps the business.

What glass is

A Rust MCP server giving an AI agent a closed build → see → interact → debug loop over external native GUI apps: launch, capture (lossless WebP), inject mouse/keyboard, read logs, diff against baselines, wait-until-stable, and read/drive the accessibility tree — served over MCP (stdio, or serve --http). Backends behind a Platform seam: X11 and Wayland (headless sway) on Linux, Windows (WGC/SendInput), Android (native apps in an AVD emulator over adb, host-OS-agnostic; clipboard + high-fidelity input via an optional on-device companion agent), iOS (native apps in the Simulator over xcrun simctl; capture, logs, clipboard, plus input and the accessibility tree via idb_companion, multi-touch gestures excepted), macOS (ScreenCaptureKit capture, CGEvent input, AXUIElement windows/logs, accessibility tree, clipboard (isolated + working under containment for apps not built with hardened runtime, via a DYLD_INSERT_LIBRARIES swizzle shim; hardened-runtime apps fall back to unsupported), sandboxing (Seatbelt)).

Layout

Cargo workspace at the repo root. Crates: glass-core (platform-agnostic heart — the Platform/Accessibility seams, session, Frame, diff, stability, log buffer), the backends (glass-x11, glass-wayland, glass-windows, glass-android, glass-ios (the iOS Simulator backend over xcrun simctl), glass-macos), the a11y readers (glass-a11y-linux, glass-a11y-windows, glass-a11y-macos; the Android uiautomator reader lives in glass-android), glass-sandbox-linux, glass-sandbox-macos, the glass-mcp server binary, and the glass-testapp fixture. glass-android also holds the host-side client + lifecycle for two optional on-device companions — an app_process agent (clipboard + high-fidelity input) and an AccessibilityService (Compose-rich a11y tree + high-fidelity set_value); both live in the separate repo github.com/fixed-width/glass-android-agent, driven over adb forward.

Read the full file on GitHub · 78 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 First seen · 78 lines · 1,112 tokens per session scan A 5869837398a9

Subscribe to this mod's changes

glass AGENTS.md is an instructions file published in the GitHub repository fixed-width/glass (12 stars, last pushed 2d ago), licensed Apache-2.0. It adds 1,112 tokens to every session, about $0.0056 per session on Opus 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 2026-08-30.