Skills agent-browser

📦

agent-browser

Name: agent-browser
Author: inference-sh-skills

Content revision r1 High Risk ⚙️ External commands🌐 Network access📁 Filesystem access

Automate Browser Sessions with Agentic Refs

AI agents often need reliable browser control without parsing full page DOMs. This skill provides compact refs, actions, screenshots, and reusable workflows.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "agent-browser" from https://skillstore.io/skills/inference-sh-skills-agent-browser.md and its manifest at https://skillstore.io/api/skills/inference-sh-skills-agent-browser/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/inference-sh-skills-agent-browser.md Signed manifest GET /api/skills/inference-sh-skills-agent-browser/manifest Signed lockfile GET /api/skills/inference-sh-skills-agent-browser/lockfile

Test it

Using "agent-browser". Inspect a product page and identify the main actions.

Expected outcome:

The page loaded successfully. The main actions are search, add to cart, size selection, and account sign in.

Using "agent-browser". Record a failing checkout path for debugging.

Expected outcome:

The workflow produced a screenshot, a short recording, and a note that the payment step failed after address entry.

Using "agent-browser". Extract links from an approved documentation page.

Expected outcome:

Captured the page title and visible navigation links.
Saved a full page screenshot for reference.
Reported stale refs after navigation and refreshed the snapshot.

Security Audit

High Risk

v4 • 7/6/2026 Open versioned report

Most static command, URL, path, and device findings are documentation or shell-template false positives. Confirmed risks include browser capture upload, SOCKS proxy support, cookie extraction guidance, sensitive session recording examples, and arbitrary page JavaScript execution. No prompt injection attempt was found in the reviewed files.

Files scanned

2,110

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (5)

High

SKILL.md:3

Screen capture upload

description: "Browser automation for AI agents via inference.sh. Navigate web pages, interact with e

The skill explicitly advertises screenshots and video recording through inference.sh. Browser captures can expose sensitive page contents when used on authenticated or private sessions.

High

references/proxy-support.md:85-123 references/proxy-support.md:145-153

Proxy Rotation for Rate-Limit Avoidance

The proxy guide includes a rate-limit avoidance workflow that rotates proxies while scraping pages. This framing can enable policy evasion and abusive automation.

The document explicitly labels a section Rate Limit Avoidance and shows proxy rotation for web scraping. The privacy proxy example confirms origin-hiding support.

High

references/authentication.md:209-224

Session Cookie Extraction Guidance

The authentication guide instructs users to extract cookies with browser JavaScript and print them. Cookies can function as session credentials.

The section is explicitly named Cookie Extraction and includes document.cookie plus echoing the resulting cookie string.

High

references/authentication.md:151-156 references/video-recording.md:189-210

Sensitive Session Recording Guidance

The docs show recording a login or admin workflow, including 2FA and administrative pages. These recordings can capture credentials, prompts, and protected data.

The examples enable record_video for login and admin workflows. The same file later warns not to record sensitive sessions, confirming the sensitivity.

Medium

SKILL.md:182-190 references/commands.md:162-170

Arbitrary JavaScript Execution in Browser Context

The skill exposes a page execute function for custom JavaScript. This can extract page data and interact with content beyond visible element refs.

The execute function is documented as running caller-provided JavaScript on the page. The examples show direct DOM reads, which is expected but sensitive on private pages.

Capability review items (6)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High

references/authentication.md:215-217

Shell command substitution

RESULT=$(belt app run agent-browser --function execute --session $SESSION --input '{

This command starts a documented cookie extraction workflow that runs JavaScript in an authenticated browser context. Session cookies are sensitive credentials.

High

references/authentication.md:218

Shell command substitution

COOKIES=$(echo $RESULT | jq -r '.result')

This command stores extracted cookie data from the browser session and the next line prints it. Exposing cookies can leak session credentials.

High

references/authentication.md:222-223

Shell command substitution

RESULT=$(belt app run agent-browser --function execute --session $SESSION --input '{

This command continues the cookie extraction section with browser-side JavaScript over session data. It can expose sensitive authenticated browsing metadata.

High

references/proxy-support.md:152

SOCKS proxy

"proxy_url": "socks5://privacy-proxy.example.com:1080"

The proxy guide documents SOCKS5 routing for browser traffic. This can hide origin traffic and support scraping or access-control bypass when misused.

High

references/proxy-support.md:165

SOCKS proxy

### SOCKS5 Proxy

The SOCKS5 section confirms support for anonymizing proxy types. In this browser automation context, that is a real abuse-enabling capability.

High

references/proxy-support.md:168

SOCKS proxy

{"proxy_url": "socks5://proxy.example.com:1080"}

The example configures a socks5 proxy URL for browser sessions. SOCKS routing can conceal automation source and bypass network controls.

Risk Factors

⚙️ External commands (164)

🌐 Network access (54)

📁 Filesystem access (15)

references/authentication.md:5 references/authentication.md:263 references/commands.md:3 references/commands.md:59 references/proxy-support.md:5 references/proxy-support.md:272 references/snapshot-refs.md:5 references/video-recording.md:5 references/video-recording.md:239 templates/authenticated-session.sh:30 templates/capture-workflow.sh:28 templates/capture-workflow.sh:78 templates/capture-workflow.sh:81 templates/capture-workflow.sh:149 templates/form-automation.sh:23

Detected Patterns

Screen capture upload

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/inference-sh-skills-agent-browser/audits/4?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/inference-sh-skills-agent-browser/security.svg)](https://skillstore.io/skills/inference-sh-skills-agent-browser?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/inference-sh-skills-agent-browser?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/inference-sh-skills-agent-browser/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/inference-sh-skills-agent-browser.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

inference-sh-skills. (2026). agent-browser security audit report (audit version 4) [Author version unspecified]. Skillstore. https://skillstore.io/skills/inference-sh-skills-agent-browser/audits/4

BibTeX citation

@techreport{inference-sh-skills-inference-sh-skills-agent-browser-2026,
  author = {inference-sh-skills},
  title = {agent-browser security audit report (audit version 4)},
  institution = {Skillstore},
  year = {2026},
  number = {4},
  url = {https://skillstore.io/skills/inference-sh-skills-agent-browser/audits/4},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "agent-browser security audit report (audit version 4)"
version: "unspecified"
type: report
authors:
  - name: "inference-sh-skills"
date-released: "2026-07-06"
url: "https://skillstore.io/skills/inference-sh-skills-agent-browser/audits/4"
identifiers:
  - type: other
    value: "skillstore:inference-sh-skills-agent-browser:audit:4"
    description: "Skillstore immutable audit report identifier"

Compare variants

19 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Highest Skillstore Score

Author / source Skillstore Score Evidence Confidence Skillstore usage Updated

supercent-io Recommended

supercent-io-agent-browser

Skillstore Score 78

Evidence Confidence High

Skillstore usage 30

Updated

2026-08-21

101-skills

101-skills-agent-browser

Skillstore Score 74

Evidence Confidence Medium

Skillstore usage 1

Updated

2026-08-21

vercel-labs

vercel-labs-agent-browser

Skillstore Score 38

Evidence Confidence High

Skillstore usage 66

Updated

2026-08-21

qu-skills

qu-skills-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 40

Updated

2026-08-21

am-will

am-will-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 36

Updated

2026-08-21

tul-sh

tul-sh-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 35

Updated

2026-08-21

inference-sh-8

inference-sh-8-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 25

Updated

2026-08-21

toolshell

toolshell-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 16

Updated

2026-08-21

skillssh

skillssh-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 16

Updated

2026-08-21

inference-sh-9

inference-sh-9-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 15

Updated

2026-08-21

inferen-sh

inferen-sh-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 10

Updated

2026-08-21

inference-shell

inference-shell-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 10

Updated

2026-08-21

inferencesh

inferencesh-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 10

Updated

2026-08-21

inference-sh-skills Current

inference-sh-skills-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 9

Updated

2026-08-21

inference-sh-0

inference-sh-0-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 9

Updated

2026-08-21

infsh-skills

infsh-skills-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 8

Updated

2026-08-21

inf-sh

inf-sh-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 8

Updated

2026-08-21

inference-skills

inference-skills-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 6

Updated

2026-08-21

halt-catch-fire

halt-catch-fire-agent-browser

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 3

Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Validate Web Flows

Run repeatable browser steps, capture evidence, and inspect failures in staging or production-like pages.

Collect Web Research

Open pages, extract visible text and links, and save screenshots for later analysis.

Prepare Support Documentation

Record browser workflows with cursor indicators and capture screenshots for customer-facing guides.

Try These Prompts

Open and Inspect a Page

Open the target page with agent-browser, list the visible interactive elements, and summarize the page purpose.

Fill a Known Form

Use the current snapshot refs to fill the contact form fields, submit it, and verify the resulting page state.

Capture Evidence for a Bug

Start a recorded browser session, reproduce the issue step by step, take a full page screenshot, and close the session.

Extract Structured Page Content

Navigate to each approved URL, refresh refs after every page change, extract visible text and links, and note any access failures.

Best Practices

Use only approved URLs and respect site terms, rate limits, and access controls.
Refresh snapshots after navigation, submissions, modals, or any dynamic page update.
Disable recording and redact screenshots when pages may contain credentials or private data.

Avoid

Do not use stale @e refs after a page changes or reloads.
Do not record login, payment, admin, or customer data workflows without explicit approval.
Do not use proxy rotation to bypass rate limits, identity controls, or geographic restrictions.