# Chrome Devtools > Inspect, debug, and automate Chrome through DevTools, including performance and accessibility. - **Type:** Skill - **Install:** `agentstack add skill-fmind-dot-chrome-devtools` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [fmind](https://agentstack.voostack.com/s/fmind) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [fmind](https://github.com/fmind) - **Source:** https://github.com/fmind/dot/tree/main/skills/chrome-devtools - **Website:** https://fmind.dev ## Install ```sh agentstack add skill-fmind-dot-chrome-devtools ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Chrome DevTools Use Chrome DevTools MCP or its experimental CLI for live browser diagnostics. [playwright](../playwright/SKILL.md) owns repeatable end-to-end journeys; [web-frontend](../web-frontend/SKILL.md) owns platform implementation guidance. ## Setup Require Node.js LTS, `npm`/`npx`, and supported Chrome. Reuse an available DevTools connection; otherwise follow [mcp-setup](../mcp-setup/SKILL.md) for the active harness. The launch command below uses a reviewed package pin; recheck installed help before updating it. ```bash npx --yes chrome-devtools-mcp@1.9.0 --isolated --headless --no-usage-statistics --no-performance-crux ``` The MCP client starts that process over stdio. Use a dedicated browser profile. Keep remote debugging on loopback; browser content, network bodies, cookies, screenshots, and traces may expose private data. The two opt-out flags disable MCP usage statistics and CrUX URL lookups respectively. For shell workflows, the same package exposes `chrome-devtools`; resolve it with `npm exec --yes --package=chrome-devtools-mcp@1.9.0 -- chrome-devtools `. Inspect `status` before starting a daemon, then explicitly start the task's session with `start --workspace="$PWD" --no-usage-statistics --no-performance-crux`. The CLI otherwise starts a persistent daemon automatically and enables unrestricted file access by default. Read `start --help` for the installed flags; do not stop or repurpose another task's daemon. ## Workflow 1. **Identify the target**: list pages and select the intended page ID, URL, viewport, and browser mode. Take a fresh accessibility snapshot before using element UIDs; navigation and rerenders can invalidate them. 1. **Reproduce the symptom**: capture console errors and relevant failed requests, then reduce to the smallest repeatable action. Redact credentials and private request data from artifacts. 1. **Measure performance**: record a bounded trace of the same action before and after a change; preserve CPU/network throttling, cache conditions, viewport, and tool versions. Stop traces you started and save artifacts in the authorized workspace. 1. **Check accessibility**: inspect roles, names, focus order, keyboard operation, and visible contrast. Combine automated checks with manual interaction; an accessibility tree or Lighthouse score alone does not establish WCAG conformance. 1. **Investigate memory and cookies**: compare repeated lifecycle actions and heap snapshots when those tools are available; inspect `HttpOnly`, `Secure`, `SameSite`, and partitioning in the request's actual context. A single heap size or cookie attribute is not a diagnosis. 1. **Verify the fix**: repeat the reproduction and relevant measurements, then add a regression test through the owning project workflow. Close task-owned pages and stop only a daemon created for this task. ## Gotchas - **Tool availability**: inspect the live tool schema or CLI help before calling a capability; optional categories, page routing, and CLI arguments vary by version. - **Browser mode**: set headed or headless explicitly for reproducibility; the CLI defaults to headless, which need not match a separately configured MCP connection. - **Lab versus field**: a local trace measures this run. CrUX field data and lab measurements describe different populations and time windows. ## Official Skills Upstream: [ChromeDevTools/chrome-devtools-mcp skills](https://github.com/ChromeDevTools/chrome-devtools-mcp/tree/main/skills). Use `skills add ChromeDevTools/chrome-devtools-mcp --list`, then follow the shared [vendor-skill policy](../agent-project/references/vendor-skills.md) for the required guidance. ## Documentation - [Chrome DevTools for agents](https://github.com/ChromeDevTools/chrome-devtools-mcp) · [CLI](https://github.com/ChromeDevTools/chrome-devtools-mcp/blob/main/docs/cli.md) · [Tool reference](https://github.com/ChromeDevTools/chrome-devtools-mcp/blob/main/docs/tool-reference.md) - Releases: [chrome-devtools-mcp](https://github.com/ChromeDevTools/chrome-devtools-mcp/releases) - Companion skills: [mcp-setup](../mcp-setup/SKILL.md), [playwright](../playwright/SKILL.md), [web-frontend](../web-frontend/SKILL.md), [benchmark](../benchmark/references/command-http.md), [quality-assurance](../quality-assurance/SKILL.md). ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [fmind](https://github.com/fmind) - **Source:** [fmind/dot](https://github.com/fmind/dot) - **License:** MIT - **Homepage:** https://fmind.dev Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-fmind-dot-chrome-devtools - Seller: https://agentstack.voostack.com/s/fmind - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.