gstack

Provides AI agents with structured workflow skills and a persistent headless Chromium browser daemon.

Updated May 15, 2026
One-click install
npx skills add https://github.com/tgmarinho/canetaco --skill gstack-tgmarinho
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gstack
Source: https://github.com/tgmarinho/canetaco/tree/main/.claude/skills/gstack
Command: npx skills add https://github.com/tgmarinho/canetaco --skill gstack-tgmarinho

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires playwright, @huggingface/transformers, and includes scripts (resource) and references (resource) components.

What problem does it solve? AI coding agents lack structured workflows for planning, reviewing, testing, and shipping software, and browser automation through agents is slow because every command cold-starts a browser. gstack gives agents a library of specialist SKILL.md workflows plus a persistent Chromium daemon that answers commands in ~100-200ms. ## Core Features & Use Cases - Workflow skill library: Plan-mode reviews (CEO, engineering, design, DX), pre-landing PR review, systematic debugging, QA with a real browser, and release/deploy skills like /ship and /land-and-deploy. - Persistent browser daemon: A compiled Bun CLI drives headless or headed Chromium over localhost HTTP with ref-based element selection, cookie import, screenshots, and ~70 commands. - Codified browser-skills: /scrape prototypes a page extraction once, /skillify converts it into a deterministic Playwright script, and repeat runs finish in ~200ms. - Use Case: A developer runs /autoplan to get CEO, design, eng, and DX reviews of a feature plan, then /qa to click through the app in a real browser, then /ship to test, push, and open the PR. ## Quick Start Ask the agent to run /office-hours to reframe your product idea, or invoke /qa to open a browser and test your running app end to end.

Frequently Asked Questions about gstack

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I give an AI agent structured code review workflows?

Install gstack and invoke skills like /review for pre-landing PR checks or /plan-eng-review for architecture review. Each skill is a SKILL.md file that loads specialist instructions into the agent's context when invoked by name.

How can an AI agent control a web browser with low latency?

gstack runs a persistent Chromium daemon that the CLI talks to over localhost HTTP. The first call starts the browser in about 3 seconds; every subsequent command completes in roughly 100-200ms with cookies and tabs preserved.

What is the difference between /scrape and /skillify in gstack?

/scrape drives a page once through agent exploration to extract data, taking around 30 seconds. /skillify codifies that successful flow into a deterministic Playwright script, so repeat scrapes of the same intent run in about 200ms.

Does gstack work on Windows?

Yes, with caveats. A curated Windows-safe test subset runs on windows-latest CI, and the setup script uses copy fallbacks instead of symlinks on MSYS2/Git Bash. Native PowerShell support is listed as a future expansion.

How does gstack prevent prompt injection from web pages?

It layers defenses: content sanitization and datamarking, a local BERT-based ML classifier scanning tool outputs, a transcript-level classifier, and a canary token that blocks the session if the system prompt leaks. Tunnel access uses scoped tokens on a separate listener.

When should I not use the gstack browser daemon?

Avoid it for one-off screenshots where a cold Playwright launch is acceptable, or when you need WebSocket streaming or MCP protocol integration, both of which gstack intentionally omits in favor of plain HTTP and stdout.