agentsclimarketplace
Sign in

Chrome

Skill litingyes/chrome-skills/skills/chrome

Runs local headless Chrome via composable cdp commands, session-based extract (article, layout, a11y), and search/fetch/audit recipes. Use when the user invokes /chrome, asks for browser CDP automation, page extraction, UI layout audit on a dev URL, Google search, or structured content from a URL.From its SKILL.md

Install
npx -y skills add litingyes/chrome-skills --skill chrome

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.6 KB, 875 tokens by cl100k_base, as published. Nobody here has run it

Chrome

Composable local Chrome automation for agents. Five top-level commands:

CommandLayerPurpose
cdpL1 atomicLaunch, navigate, wait, evaluate, snapshot, emulate, screenshot, close
extractL2 composeExtract structured data from the current page (article, serp, layout, a11y)
searchL3 recipeGoogle search one-shot
fetchL3 recipeSingle-URL article extraction one-shot
auditL3 recipeMulti-viewport UI layout verification on a dev URL

Prerequisites

  • Google Chrome or Chromium installed locally
  • Python 3.10+
  • Install once:
skills/chrome/scripts/setup

This installs Python dependencies into skills/chrome/scripts/vendor/.

Optional: set CHROME_PATH to override Chrome binary discovery.

Routing

  1. Parse the first token after /chrome: cdp, extract, search, fetch, or audit.
  2. Read only skills/chrome/references/<token>.md for workflow details.
  3. Run the matching script under skills/chrome/scripts/.
  4. Scripts print JSON to stdout. Summarize for the user; keep the JSON for reasoning.

Script paths (run from repo root):

skills/chrome/scripts/cdp <action> [options]
skills/chrome/scripts/extract <mode> --session <id>
skills/chrome/scripts/search "<query>"
skills/chrome/scripts/fetch "<url>"
skills/chrome/scripts/audit "<url>" [options]

When to use which layer

  • One URL, article content → fetch
  • Google search → search
  • UI layout on dev server → audit (or extract layout in a session)
  • Multiple pages in one browser → cdp launch → navigate + extract (repeat) → cdp close
  • Custom JS or debugging → cdp evaluate
  • CAPTCHA / bot check on Google → see references/search.md; retry with cdp launch --headed --user-data-dir <profile>
  • Core Web Vitals / CLS → future extract perf (see extension-roadmap.md)

UI verification principles

  • Never trust CSS alone — verify rendered layout with audit or extract layout
  • Always multi-viewport — use audit --viewports 375,1280 (mobile + desktop)
  • Issues are facts — fix from issues[] (rule, selector, viewport), not HTML guesses
  • Polish layer — audit --include-screenshot then visual review
  • Close sessions — audit auto-closes; manual flows need cdp close

Command index

cdp actions

launch · navigate · wait · evaluate · snapshot · emulate · resize · screenshot · close

Details: references/cdp.md

extract modes

article · serp · layout · a11y

Requires --session. Details: references/extract.md

Recipes

JSON schemas: references/output-schemas.md

Additional resources

What ships with it: 26 files

85.5 KB alongside SKILL.md, 16 of them executable

references/

scripts/

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.