Back to skill

Security audit

账号切换器

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed local controller for a browser account management tool, with some installation and safety caveats but no evidence of hidden or malicious behavior.

Install this only if you trust the third-party multi-account management application and understand that start/stop commands can affect active browser account sessions. Use it with explicit account names or indexes, and consider installing Python dependencies in a virtual environment.

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:81
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 81-91
Vulnerability Type: Unverified and unpinned third-party dependencies
Risk Level: Medium

Vulnerable Code Snippet

bash
pip install requests

The same installation section directs users to obtain a proprietary supporting application from:

text
https://zmt.scys6688.com/

Technical Analysis

The installation instructions use pip install requests without specifying an approved version, package hash, lockfile, or trusted package index. Consequently, the dependency resolved at installation time can differ from the version reviewed during the audit.

The instructions also direct users to obtain a required proprietary application from an external website without documenting an expected version, cryptographic checksum, digital signature, or verification procedure. The effective code installed by users therefore depends on mutable external supply-chain sources.

This does not establish that the current requests package or external application is malicious. The risk arises if a package release, publisher account, package repository, website, DNS route, or downloadable installer is compromised after the Skill has been reviewed.

Attack Path

  1. An attacker compromises a dependency publisher account, package distribution channel, external website, or installer delivery infrastructure.
  2. The attacker publishes or distributes a modified package or installer containing malicious installation or runtime behavior.
  3. A user follows the Skill documentation and runs the unpinned installation command or downloads the unverified supporting application.
  4. The package installer, imported dependency, or external application executes the attacker's code.
  5. The malicious code operates with the privileges of the user performing the installation or running the Skill.

Impact Assessment

Successful exploitation could permit arbitrary ...[truncated 575 chars]

Remediation
View remediation

Remediation Suggestions

  1. Declare Python dependencies in a reviewed requirements or lock file using an exact version.
  2. Require cryptographic hashes, for example with pip install --require-hashes -r requirements.txt.
  3. Explicitly use a trusted package index and prevent fallback to unexpected indexes.
  4. Test and periodically review pinned dependency versions for known vulnerabilities.
  5. Document the exact expected version of the proprietary supporting application.
  6. Publish a fixed HTTPS download location together with an expected SHA-256 checksum and, where available, a verifiable publisher signature.
  7. Instruct users to verify the checksum and digital signature before executing the installer.
  8. Recommend installation in an isolated virtual environment under a non-administrative account.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill explicitly supports starting and stopping browser accounts, which are state-changing operations that can disrupt active sessions, terminate user workflows, or alter account state without any embedded warning, confirmation, or safety guidance. In this context, the capability is legitimate, but the absence of user-facing safeguards makes accidental or unauthorized disruption more likely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The module docstring and all user-facing messages are written exclusively in Chinese, including usage instructions and error text. For a general-purpose account-management tool, this imposes a specific language on users without any opt-in, fallback, or documented regional justification.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The module docstring states that execution results are both printed to the console and written to last_result.txt. However, for start/stop operations the code prints both a success line and the API response to stdout, while only result_text is persisted via write_result, so the file does not contain the full execution output that was printed.

Content

No source excerpt is available for this finding.