Dev-Browser skill
Automate Chrome with persistent named pages, Puppeteer scripts and a local command-line service.
Community · Sawyer Hood| Author | Sawyer Hood |
|---|---|
| Licence | MIT |
| Source | SawyerHood/dev-browser · SKILL.md on GitHub |
| Works in | Claude Code · Codex, Gemini CLI, Cursor, Copilot |
| Read from the source on |
The dev-browser skill guides an agent through real Chrome sessions using short Puppeteer scripts and persistent named pages.
The current version uses a local CLI and a warm Bun daemon. It can navigate, fill forms, extract page data and capture screenshots; this site only documents the skill.
From the skill's SKILL.md: # dev-browser
What it does
- Inspect an interactive page snapshot before acting.
- Run a short script against a named page.
- Wait for the required state and inspect the result.
What you get
- Browser actions, extracted data or saved screenshots, depending on the task.
When to use it
- Repeat a browser workflow while keeping page state between calls.
When not to use it
- Windows or musl Linux, which the current release does not support.
What it needs
- A shell-capable agent, supported macOS or glibc Linux, and Chrome.
- Installation may download a runtime binary and Chrome. Agent usage and external sites may have separate costs or accounts.
Files in the skill
| File or folder | What it is for |
|---|---|
SKILL.md | CLI workflow, named pages, element references and navigation cautions. |
Licence as found in the repository on 2026-11-20: MIT.
Example prompts
Type one of these after installing; naming the skill makes sure it is used.
- Inspect the checkout form and report its required fields without submitting it.
- Capture desktop and phone screenshots of the local preview.
- Reuse the named page to verify that the filter changed the results.
How to install dev-browser
CLI setup and browser permissions
Install the CLI separately from the instruction folder. Download Chrome only if the CLI reports that none is available; read its current help before using it.
Review scripts and authorize browser actions. Profiles can retain cookies and logins; attaching to an existing browser can expose its active sessions.
The runtime guide is in the repository; it is not bundled inside this skill folder.
npm install -g dev-browser
# Only if the CLI reports "No Chrome found":
dev-browser install
dev-browser --helpA skill is a folder with a SKILL.md file. You install it once; the agent reads its short description at the start of each session and loads the full instructions only when a request matches. Pick the place you use:
Claude Code: copy the folder into your skills folder
Put the skill's folder in ~/.claude/skills/ to use it in every project on this computer, or in .claude/skills/ inside a repository to share it with everyone who works there (commit it).
git clone --depth 1 https://github.com/SawyerHood/dev-browser.git
mkdir -p ~/.claude/skills
cp -R dev-browser/skills/dev-browser ~/.claude/skills/Claude Code notices new or changed skills in these folders during the session. If the skills folder did not exist when the session started, run /reload-skills once. Then type /dev-browser or just describe the task.
On Windows, run the commands in Git Bash or WSL, or copy the folder into the same place with File Explorer.
Codex, Gemini CLI, Cursor and GitHub Copilot
These agents read the same SKILL.md folders. The one folder all four accept inside a project is .agents/skills/, so this copies the skill there:
git clone --depth 1 https://github.com/SawyerHood/dev-browser.git
mkdir -p .agents/skills
cp -R dev-browser/skills/dev-browser .agents/skills/| Agent | Project folder | Your user folder | Notes |
|---|---|---|---|
| OpenAI Codex | .agents/skills/ | ~/.agents/skills/ | Type $ or run /skills to pick a skill; Codex also chooses one when your task matches. |
| Gemini CLI | .agents/skills/.gemini/skills/ | ~/.agents/skills/~/.gemini/skills/ | Or install from Git with gemini skills install <git URL> --path <folder> --consent. Google replaced Gemini CLI with Antigravity CLI for free-tier and Google One users on 18 June 2026; Antigravity uses .agents/skills/ in a project and ~/.gemini/config/skills/ for you. |
| Cursor | .agents/skills/.cursor/skills/ | ~/.agents/skills/~/.cursor/skills/ | Type / in Agent chat to pick a skill. Cursor also reads Claude Code's .claude/skills/ folders. |
| GitHub Copilot | .github/skills/.agents/skills/.claude/skills/ | ~/.copilot/skills/~/.agents/skills/ | Works in Copilot CLI, the cloud agent, and agent mode in VS Code, Visual Studio and JetBrains (IDE support depends on the version). Or run gh skill install OWNER/REPOSITORY SKILL (public preview, GitHub CLI 2.90.0 or later). |
Sources for these steps
Checked against each vendor's documentation on 2026-11-14:
Persistent pages and fresh references
Page names are scoped to a browser. Headed and headless modes use separate Chrome profiles; do not issue concurrent calls against the same named page.
Element references reset after navigation. Take a new snapshot and wait for the required element or load state rather than assuming a click finished navigation.
Installation choices and limits
The publisher also provides an install-skill command for supported agent folders. Review its targets before running it; copying this folder alone does not install dependencies.
The npm package obtains a bundled runtime binary. Normal use does not require a separate Node runtime afterward. Keep an attached, user-owned browser open when finishing the task.
Questions about dev-browser
Does copying SKILL.md install the runtime?
No. The CLI and a usable Chrome installation are separate requirements.
Does it still use the old Playwright server workflow?
The checked version uses Puppeteer through the dev-browser CLI. Older server setup instructions may no longer apply.
Is an MCP server required?
No. The CLI works directly; the repository also documents an optional MCP mode.
Is script execution a security sandbox?
No. The documented node:vm isolation is not a security boundary. Run only reviewed scripts with suitable permissions.
More AI skills
Related skills
Agent Evaluation skill
Community · Muratcan KoylanPlan agent regression suites with deterministic checks, separate quality dimensions and versioned baselines.Tool Design skill
Community · Muratcan KoylanDesign agent-facing tool contracts with clear inputs, outputs, selection guidance and recoverable errors.Context Degradation skill
Community · Muratcan KoylanDiagnose long-context failures and plan recovery using verified facts and workload-specific comparisons.Advanced Evaluation skill
Community · Muratcan KoylanDesign LLM judges with evidence-based rubrics, swapped pairwise comparisons and human calibration.Brainstorming skill
Community · Jesse Vincent / Prime RadiantTurn an undeveloped idea into an agreed design before writing implementation code.Writing Plans skill
Community · Jesse Vincent / Prime RadiantConvert an approved design into reviewable engineering tasks with interfaces and verification.