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.
npx skills add sderosiaux/chrome-agent --skill scrape-structured-datagit clone --depth 1 https://github.com/sderosiaux/chrome-agentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/sderosiaux/chrome-agent/scrape-structured-data)<a href="https://agentmods.dev/skills/sderosiaux/chrome-agent/scrape-structured-data"><img src="https://agentmods.dev/badge/skills/sderosiaux/chrome-agent/scrape-structured-data/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/sderosiaux/chrome-agent/scrape-structured-data"><img src="https://agentmods.dev/badge/skills/sderosiaux/chrome-agent/scrape-structured-data.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high YARA Match · line 81 YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).Fix: Remove the malware payload or compromised file entirely. Investigate how it entered the skill and audit all other artifacts for additional indicators of compromise.
- high Privilege Escalation · line 82 Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.Fix: Remove references to credential paths. Use environment variables or secrets managers. For docs, use placeholder paths (e.g., /path/to/config). Never load .env or token files in production code paths.
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00111 | $0.01147 |
| Opus 5.5 | $0.00044 | $0.00459 |
| Sonnet 5.5 | $0.00022 | $0.00229 |
| Haiku 4.5 | $0.00011 | $0.00115 |
Grade A, and why
scrape-structured-data 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 1mo ago.
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.
How it starts
The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scrape structured data from a page
chrome-agent extract finds the repeating record pattern on a page and returns it as JSON. No
selectors to write, no HTML read into the model to find them.
which chrome-agent || npm install -g chrome-agent
chrome-agent goto https://news.ycombinator.com
chrome-agent --json extract --limit 50
{"count":30,"pattern":"TR","items":[
{"title":"PGSimCity - How PostgreSQL Works","url":"https://nikolays.github.io/PGSimCity/","fields":["..."]},
...
]}
Check count against what the page shows. Wildly off means the page has more than one repeating
pattern — scope it with --selector.
Why extract rather than read the page
Measured on the Hacker News front page. All three contain the same 30 stories; only the first hands them over as records.
| approach | tokens | what you get |
|---|---|---|
extract --limit 30 |
~1,570 | all 30 stories as records with URLs |
inspect (accessibility tree) |
~5,650 | the tree, stories mixed into the page chrome |
| raw HTML | ~8,730 | everything, including markup you'll never read |
3.6x against the tree, 5.6x against raw HTML here. The margin narrows on pages that are nothing but a list: on a blog archive, ~12,500 tokens against ~16,100 for the tree. The win comes from pages where the records sit inside a lot of other markup.
Scoping, lazy loading, infinite scroll
chrome-agent extract --selector ".product-grid" # restrict to one container
chrome-agent extract --limit 100 # default is 10
chrome-agent extract --scroll # scrolls, waits for new nodes, then extracts
chrome-agent extract --a11y --scroll --limit 50 # React/SPA feeds where the DOM is noise
--a11y reads the accessibility tree instead of the DOM. Use it when a site renders into nested
generated <div>s (X.com, most React apps) and plain extract returns junk.
When extract returns nothing
A page with no repeating pattern gives an empty list and a hint — the correct answer for an article or a landing page. Fall back in this order:
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.
- 1mo ago Changed · -24 lines 778fef5f7259
- 1mo ago First seen · 127 lines · 111 tokens per session scan A 5c3a5c678657
scrape-structured-data is a skill published in the GitHub repository sderosiaux/chrome-agent (95 stars, last pushed 10d ago), licensed MIT. It adds 111 tokens to every session and 1,147 once invoked, about $0.0004 per session on Opus 5.5. 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.
Other skills, from other repositories
owb
Open Web Bridge (OWB) — drive the user's own real browser with the owb command. Read pages behind their existing logins, gather and cross-check information, fill forms, walk multi-step flows, debug their site, audit responsive/accessibility behavior, and capture or reverse-engineer network traffic. Use this whenever…
cdpilot
Browser automation over raw Chrome DevTools Protocol — navigate, click, fill forms, screenshot, extract content, handle iframes, and bypass basic bot detection. Use whenever a task needs a real browser (no Puppeteer/Playwright/Selenium available, or when those get detected) — for example "open this page and click X"…
map
Per-site navigation memory for tandem's shared browser. Profiles that describe a site's skeleton (routes), its durable locators, navigation recipes, and gotchas, so you navigate KNOWING instead of re-deriving the DOM each time. Use it when you start operating a site (check whether a profile already exists), when you…
codex-control-chrome-mcp
Use Codex Control Chrome MCP to control the user's normal Chrome profile through the Codex Chrome extension. Use when a task needs existing Chrome tabs, cookies, logged-in sessions, installed extensions, CDP commands, page inspection, navigation, screenshots, network/console events, or browser automation from an MCP…
tandem
How to operate the Chrome browser shared with the human (the tandem MCP's browser tools). Use it whenever you need to navigate, read/analyze pages, get past blocks that require human interaction (captchas, anti-bot checkpoints, logins), or inspect the network/DOM in real time alongside the human.
verify-chrome-cdp-ex
Verify chrome-cdp-ex (CLI and MCP driving a real Chrome). Use when proving a fix, or when a reviewer re-runs the user path on a PR head.