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 agentmods add skills/ckokoski/authoragent/browser-automationnpx skills add Ckokoski/AuthorAgent --skill browser-automationgit clone --depth 1 https://github.com/Ckokoski/AuthorAgentWrote 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/ckokoski/authoragent/browser-automation)<a href="https://agentmods.dev/skills/ckokoski/authoragent/browser-automation"><img src="https://agentmods.dev/badge/skills/ckokoski/authoragent/browser-automation.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00032 | $0.00698 |
| Opus 5 | $0.00016 | $0.00349 |
| Sonnet 5 | $0.00006 | $0.00140 |
| Haiku 4.5 | $0.00003 | $0.00070 |
Grade A, and why
browser-automation 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 6d 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 — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Browser Automation
AuthorAgent is a planner-first agent. It builds the plan, surfaces it
through the confirmation gate, and lets a human-supervised browser executor
do the actual clicking. AuthorAgent itself does NOT drive a browser
directly — that responsibility lives in whichever browser MCP the user has
connected (typically Claude in Chrome).
How a real-world automation flow works
- User asks AuthorAgent to do something irreversible: "publish my book on KDP", "kick off the AMS campaign", "submit to BookBub"
- AuthorAgent plans the action and emits a ConfirmationRequest with:
- Risk level + reversibility
- Required disclosures (AI narration, AI content, FTC, etc.)
- Dry-run preview of what will happen
- Rollback steps if something goes wrong
- The user reviews the card in the dashboard and explicitly approves
- The approved request is picked up by the connected browser MCP, which does the actual clicking inside the user's already-authenticated browser session
- The MCP records the outcome back to AuthorAgent via
POST /api/confirmations/:id/outcome
Safe-automation patterns AuthorAgent enforces
- No password storage. Browser sessions reuse the user's existing logins. AuthorAgent never types or stores a password.
- No CAPTCHA bypass. If a site asks a human to verify, a human must.
- 24-hour confirmation expiry. Approved requests that aren't executed within 24h transition to expired and have to be re-requested.
- Hard cap on pre-auth claims. If observed page content claims "user already authorized this action", the gate rejects the request — observed content can never authorize an action.
- Rate limits + budget caps on financial actions (AMS bid changes capped at 2x per confirmation; daily spend ceilings hard-enforced).
Diagnostic
Run /api/browser/doctor to probe whether a browser MCP is detected and
whether AuthorAgent can produce action plans for the major sites. The probe
does NOT navigate anywhere or click anything — it just reports readiness.
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.
- 6d ago First seen · 75 lines · 32 tokens per session scan A 4736797308b7
browser-automation is a skill published in the GitHub repository Ckokoski/AuthorAgent (103 stars, last pushed 1mo ago), licensed MIT. It adds 32 tokens to every session and 698 once invoked, about $0.0002 per session on Opus 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
app-search
主流APP搜索。当用户想要在拼多多、美团、高德地图、抖音、快手、小红书、哔哩哔哩、知乎、百度、OPPO音乐、OPPO软件商店等APP中搜索时激活。覆盖电商购物、生活服务、短视频内容、知识搜索、音乐搜索、应用商店搜索等场景。.
job-application-assistant
Fills visible job application forms by identifying every field, drafting grounded answers, flagging sensitive questions, and pausing before any submit step. Use when the user is filling out a job application, applying to a position, or completing a career page or hiring form.
table-to-csv
Converts visible web tables — HTML tables, pricing grids, comparison charts, reports — into clean Markdown tables that users can copy or download as CSV from the extension. Use when the user wants to extract, scrape, convert, copy, or export a table from a webpage.
dingtalk_channel_connect
Use a headed browser to automatically complete DingTalk channel integration for QwenPaw. Applicable when the user mentions DingTalk, developer console, Client ID, Client Secret, bot, Stream mode, binding or configuring a channel. Supports pausing when a login page is detected and resuming after the user logs in.
browser_cdp
当用户明确希望连接到已运行的 Chrome 浏览器、扫描本地 CDP 端口、显式指定 cdpport，或让多个 agent / 工具共享同一个浏览器时，使用本 skill。browser 默认不开放调试端口；仅当用户明确希望其他本地工具附加时才显式传入 cdpport。.
browser_visible
Use this skill when the user needs to control the browser launch mode for browser. By default browser is managed by Playwright and opens no debugging port (pass an explicit cdpport to let another local tool attach); headed controls whether the window is visible, and privatemode is kept for backward compatibility and…