Versioned security assessment

Report ID: SA-D71C7417

7/17/2026, 10:54:25 PM

agent-browser security assessment v5

Skill Security Certification Report

Audit History
Scanner version 3.0.0 Audit model: codex Historical report
Skill name
agent-browser
Version
v5
Maintainer
101-skills
Source
github.com/101-skills/skills/tree/main/tools/utilities/agent-browser/
Coverage
9 Files scanned · 2,110 Lines analyzed
Policy version
skillstore-security-audit-policy-v1

Highest confirmed finding severity

High

6 confirmed security findings require attention.

Installation context

Historical evidence

This report may not describe the currently installable artifact. Open the current Skill page for install guidance.

Open current Skill page

This report does not block or authorize the manifest or ZIP.

Most static alerts are false positives caused by Markdown syntax, example URLs, relative links, and ordinary shell substitutions. Confirmed risks include hosted screen capture, SOCKS proxies, rate-limit evasion, secret handling, external dependency trust, and unrestricted browser actions. Publication should require stronger consent, trust, and abuse-prevention guidance.

Report position

Historical report

Open audit history before using this report to install.

Audit attestation

Attestation unavailable

No public attestation is available for this report.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

9 Files scanned · 2,110 Lines analyzed

10 items shown for review

Limitations

This report does not claim runtime or sandbox execution and does not prove the absence of side effects.

Evidence chain

Follow the evidence from source binding to the install contract. Available evidence supports verification; it is not a safety guarantee.

  1. Source

    Commit and path bound

  2. Artifact

    Content and tree hashes bound

  3. Audit

    Complete

  4. Install contract

    Open manifest to verify

    Open manifest

Capabilities observed

Observed means this report recorded supporting evidence. Not recorded does not prove that a capability is absent.

Contains scripts

May execute code included with the Skill.

Not recorded by this audit

Network access

May connect to external services.

Observed in 50 evidence locations

Filesystem access

May read or write local files.

Observed in 15 evidence locations

Env variables

May read values from the process environment.

Not recorded by this audit

External commands

May invoke commands or programs outside the Skill.

Observed in 50 evidence locations

Capability review items (4)
High
references/proxy-support.md:152
SOCKS proxy
"proxy_url": "socks5://privacy-proxy.example.com:1080"
The documentation intentionally enables SOCKS5 traffic anonymization. Combined with scraping and proxy-rotation guidance, this can bypass network or service controls.
High
references/proxy-support.md:165
SOCKS proxy
### SOCKS5 Proxy
The documentation intentionally enables SOCKS5 traffic anonymization. Combined with scraping and proxy-rotation guidance, this can bypass network or service controls.
High
references/proxy-support.md:168
SOCKS proxy
{"proxy_url": "socks5://proxy.example.com:1080"}
The documentation intentionally enables SOCKS5 traffic anonymization. Combined with scraping and proxy-rotation guidance, this can bypass network or service controls.
Low
SKILL.md:13
Hardcoded URL
![Agentic Browser](https://cloud.inference.sh/app/files/u/4mg21r6ta37mpaz6ktzwtt8krr/01kgjw8atdxgkrs
The Markdown image embeds an externally hosted inference.sh asset, so rendering can issue an unsolicited request and disclose viewer network metadata. This is a low privacy risk.

Risk findings

Confirmed security concerns are separated from items that still need review.

Confirmed security concerns (6)

RISK-001 High
SKILL.md:3
Screen capture upload
description: "Browser automation for AI agents via inference.sh. Navigate web pages, interact with e
The skill intentionally captures screenshots and videos through a hosted inference.sh browser. Captured pages may contain credentials or personal data outside the local trust boundary.
RISK-002 High
references/proxy-support.md:85-128
Rate-Limit Evasion Guidance
Documentation explicitly recommends rotating proxies for rate-limit avoidance during scraping. This facilitates bypassing service controls and can support abusive collection.
The section is titled Rate Limit Avoidance and provides a complete proxy-rotation loop for scraping multiple pages.
RISK-003 High
references/authentication.md:31-38references/authentication.md:99-102references/authentication.md:128-138references/authentication.md:209-219
Hosted Handling of Authentication Secrets
Examples send passwords and TOTP codes through hosted browser commands, then extract and print session cookies. This can expose account secrets outside the local environment.
The examples explicitly pass password and TOTP values to agent-browser and print document cookies to standard output.
RISK-004 High
references/snapshot-refs.md:24-29references/snapshot-refs.md:147-157
Untrusted Web Content Reaches Agent Context
Snapshots convert remote page text into compact input for AI decisions without warning that pages may contain malicious instructions. This creates a prompt-injection surface.
The guide explicitly positions elements_text as input for AI decisions but provides no boundary for instructions embedded in remote page content.
RISK-005 High
SKILL.md:73-91SKILL.md:158-189
Consequential Browser Actions Without Guardrails
The skill enables submissions, local file uploads, arbitrary page JavaScript, proxying, and authenticated browsing. These actions can disclose data or change accounts.
The action table and examples explicitly document file upload, form interaction, navigation, and arbitrary JavaScript without confirmation or domain restrictions.
RISK-006 Medium
SKILL.md:7SKILL.md:17
Unpinned External CLI Dependency
The skill directs users to install an external CLI skill and follow remote installation instructions without a pinned version. Upstream changes could alter executed behavior.
Both installation references resolve mutable remote content, and neither instruction pins a reviewed version or integrity digest.

Remediation

Suggested fixes recorded by this audit. Applying them is the maintainer’s responsibility.

  1. FIX-001
    High
    The proxy guide presents rotation as a method for rate-limit avoidance.
    Remove this use case and require compliance with target terms, robots directives, rate limits, and explicit authorization.
  2. FIX-002
    High
    Authentication examples send passwords and TOTP codes through the hosted browser and print cookies.
    Add a clear trust-boundary warning, prohibit cookie logging, and recommend short-lived test accounts without production privileges.
  3. FIX-003
    High
    Remote page text is presented to AI agents without prompt-injection guidance.
    State that webpage content is untrusted data and require confirmation before consequential actions requested by page text.
  4. FIX-004
    High
    Uploads, form submissions, arbitrary JavaScript, and account changes have no required confirmation step.
    Require domain allowlists and explicit user approval before uploads, submissions, purchases, messages, or account changes.
  5. FIX-005
    High
    Screenshots and videos can capture credentials, personal data, and authenticated pages.
    Document storage, retention, deletion, and access controls. Disable recording by default and require consent for sensitive sessions.
  6. FIX-006
    Medium
    The required external CLI skill and installation instructions are not version pinned.
    Pin reviewed dependency versions and publish integrity information for installation content and hosted service components.

Expert evidence

Immutable subject identity, scanner metadata, dismissed matches, and source-level evidence.

Artifact subject

Marketplace commit
d71c7417a35d5c2624161bd2fe8de8a41a362128
Content hash
433ed471ceca49f696ba72856894860594a5dccbcac82da61797ae7a3a3ebfef
Tree hash
efb3a5c2872f1f029609b8b1b25768a3862f787750acdbbc6626ac4a262e5cae
Skill path
skills/101-skills/agent-browser
Audit payload hash
e42b229177f40a7763d7e2760f53df41

Analysis metadata

Audit model: codex

Analysis state: Complete

Scope is limited to the recorded files, lines, methods, and evidence. No runtime or sandbox execution is claimed.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest Source

The manifest and lockfile bind install artifacts to cryptographic hashes. This integrity claim is separate from the security assessment.

Audit attestation: unavailable