Back to skill

Security audit

skill test

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward browser automation guide, but users should handle cookies, credentials, and saved browser state carefully.

Install from a trusted, pinned release when possible. Do not save auth.json for sensitive accounts unless you can protect and delete it like a password, and avoid printing or sharing cookies, localStorage, request headers, or captured network data.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:15
Finding

Unpinned Third-Party Package and Source Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 15-25
Vulnerability Type: Unpinned and mutable third-party dependencies
Risk Level: Medium

Vulnerable Code

bash
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
bash
git clone https://github.com/vercel-labs/agent-browser
cd agent-browser
pnpm install
pnpm build
agent-browser install

Technical Analysis

The installation instructions globally install agent-browser without pinning an audited version or verifying package integrity. The alternative source-based procedure clones the repository's mutable default branch without selecting a reviewed commit or release tag, then installs dependencies and executes a build.

Both installation paths may execute package lifecycle scripts and other upstream-controlled installation logic. Consequently, the code that runs during installation can differ from the content reviewed when this skill was published. The --with-deps option may additionally invoke operating-system dependency installation, increasing the potential effect where the environment grants package-manager or administrative privileges.

This is a supply-chain weakness rather than evidence that the named upstream package is currently malicious.

Attack Path

  1. An attacker compromises the npm package, an upstream maintainer account, the source repository, or a transitive dependency.
  2. The attacker publishes malicious package content or modifies the repository's mutable default branch.
  3. A user follows the documented global npm installation or source-build procedure.
  4. npm or pnpm retrieves the attacker-controlled content and may execute its lifecycle, build, or installation scripts.
  5. The malicious code executes with the permissions of the invoking user.
  6. If installation is performed with administrative privileges or system dependency installation receives elevated authorization, the compromise may extend to system-leve ...[truncated 725 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin agent-browser to a specifically reviewed version instead of installing the latest release:
    bash
    npm install --global agent-browser@<reviewed-version>
    
  2. Pin source installations to a verified commit hash or signed release tag:
    bash
    git clone https://github.com/vercel-labs/agent-browser
    cd agent-browser
    git checkout --detach <verified-commit-hash>
    
  3. Publish expected package integrity hashes or signature-verification instructions and require verification before installation.
  4. Use a committed lockfile with integrity metadata for source builds, and require frozen or immutable dependency installation.
  5. Prefer a project-local installation in an isolated container or restricted environment over a global installation.
  6. Review package lifecycle scripts, build scripts, and transitive dependencies before execution. Where supported, initially install with lifecycle scripts disabled and explicitly run only audited setup steps.
  7. Avoid running npm, pnpm, or the browser installer as root. Clearly document when --with-deps may request elevated privileges and provide a separately reviewed list of required operating-system packages.
  8. Periodically re-audit the pinned release and update the pin only after reviewing upstream and dependency changes.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
85% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · SKILL.md (reported line 41)May include surrounding context.

agent-browser open # Navigate to page agent-browser snapshot -i # Get interactive elements with refs agent-browser click @e1 # Click element by ref agent-browser fill @e2 "text" # Fill input by ref agent-browser close # Close browser

text

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The documented features for custom headers, HTTP basic credentials, request interception, and request inspection can expose sensitive authentication material or transmit it to unintended destinations if misused. Because the skill presents these as straightforward commands without safety notes, an agent may capture or inject secrets into traffic, increasing risk of credential leakage and request tampering.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill exposes commands to read, set, and clear cookies and localStorage without any caution about handling secrets, PII, or session material. In an automation skill used by agents, these primitives can directly access or modify authentication state and private data, making accidental disclosure or malicious misuse more likely.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly documents saving and reusing authenticated browser state (state save/load auth.json) but does not warn that the saved file can contain live session cookies, tokens, and other authentication artifacts. In an agent context, this increases the chance that credentials are persisted insecurely, reused across tasks, or exfiltrated from disk, enabling account takeover if the state file is accessed by an attacker.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

The documented state save auth.json / state load auth.json workflow enables persistence of authenticated session material across runs. Without warnings or guardrails, this can leave reusable authentication artifacts on disk and make lateral reuse or theft of sessions significantly easier in shared or multi-tenant environments.

Content

Scanner excerpt · SKILL.md (reported line 218)May include surrounding context.

bash
agent-browser state save auth.json    # Save session state
agent-browser state load auth.json    # Load saved state

Example: Form submission

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

The authentication example normalizes loading previously saved browser state into later sessions, which can encourage insecure long-lived reuse of session tokens. In practice, this can bypass reauthentication and magnify the impact of a leaked auth.json file, especially for privileged dashboards or enterprise apps.

Content

Scanner excerpt · SKILL.md (reported line 247)May include surrounding context.

agent-browser wait --url "/dashboard" agent-browser state save auth.json

Later sessions: load saved state

agent-browser state load auth.json agent-browser open https://app.example.com/dashboard

text