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 johnkozaris/web-interact-plugin --skill web-interactgit clone --depth 1 https://github.com/johnkozaris/web-interact-pluginWrote 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/johnkozaris/web-interact-plugin/web-interact)<a href="https://agentmods.dev/skills/johnkozaris/web-interact-plugin/web-interact"><img src="https://agentmods.dev/badge/skills/johnkozaris/web-interact-plugin/web-interact/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/johnkozaris/web-interact-plugin/web-interact"><img src="https://agentmods.dev/badge/skills/johnkozaris/web-interact-plugin/web-interact.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00144 | $0.05330 |
| Opus 5.5 | $0.00058 | $0.02132 |
| Sonnet 5.5 | $0.00029 | $0.01066 |
| Haiku 4.5 | $0.00014 | $0.00533 |
Grade C, and why
web-interact scanned grade C with 2 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/johnkozaris/web-interact/releases/latest/download/web-interact-installer.sh | sh Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/johnkozaris/web-interact/releases/latest/download/web-interact-installer.sh | sh How it starts
The opening of the file, as written. The whole thing — 493 lines — stays where its author put it; the contents beside it link to each section on GitHub.
web-interact
Browser automation CLI. Each shell command maps to one browser action.
Uses Playwright by default (switch to Patchright with web-interact mode assistant).
Designed for automating your own web applications — please use responsibly.
Core loop
The workflow is always: open → discover → act by index → re-discover after changes.
web-interact --headless open "https://example.com"
web-interact --headless discover # → [1] input "Email" [2] input[password] ...
web-interact --headless fill 1 "[email protected]"
web-interact --headless fill 2 "password"
web-interact --headless click 3 # submit — triggers navigation
web-interact --headless discover # ALWAYS re-discover after navigation or form submission
Re-discover after: navigation, form submission, AJAX updates, tab switches, or any action that changes the page. Indices auto-refresh when the URL changes, but explicitly re-discover after same-page DOM changes (modals, accordions, dynamic content).
Output contract
- Actions (click, fill, press, scroll, etc.): no output on success (exit 0). Error text on stderr (exit 1).
- Getters (get url, get title, eval): raw value on stdout.
- Data (tab list, cookies get, storage local): JSON on stdout.
- Screenshots: file path on stdout.
- Vision (
--vision): plain screenshot path on stderr asvision:/path/file.png. - Vision + annotate (
--vision --annotate): annotated screenshot with element overlays.
Key distinctions
fillclears existing value then types. Use for form fields.typeappends to existing value. Use for search boxes with autocomplete, or adding to existing text.clickby index uses CDP mouse events (works on any element including React/Vue/Canvas).clickby CSS selector uses JS DOM click (simpler, works for standard elements).keyboard type "text"types without targeting an element (for canvas apps, active element input).mouse click 100 200clicks at viewport coordinates (for canvas/WebGL apps where elements aren't in DOM).
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 First seen · 493 lines · 144 tokens per session scan C 91971c3af82d
web-interact is a skill published in the GitHub repository johnkozaris/web-interact-plugin (2 stars, last pushed 6mo ago), licensed MIT. It adds 144 tokens to every session and 5,330 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
firefox-control
Use to automate web browsing via Firefox WebDriver BiDi — tab listing, JS evaluation, screenshots, DOM helpers.
chrome-control
Use when debugging web applications, diagnosing page errors, inspecting console output, or capturing screenshots via Chrome DevTools Protocol (CDP).
fanqie-publisher
Publish prepared novel chapters from local Markdown files to the Fanqie Novel writer web backend via browser automation. Use when the user wants to upload chapters, continue publishing unpublished chapters, save or reuse login state, or schedule Fanqie chapter releases from a local directory.
xiaohongshu-ops
A browser-based operations assistant for Xiaohongshu, a Chinese social-media and shopping platform. It uses a logged-in Chrome browser to analyse accounts and content and manage publishing workflows.
platform-data-explorer
A platform-exploration skill that examines a website or online service to find useful browser actions and data-extraction tasks that could become skills.
browser-use
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, or extract information from web pages.