Back to skill

Security audit

B站单视频章节笔记

Security checks for vulnerabilities and agentic risk

Overview

The video-note workflow is limited, but conflicting instructions leave unclear who chooses the signed-in browser environment.

Before installing, resolve the browser-selection conflict: require your explicit choice of browserId before any page access or environment substitution. Prefer a dedicated browser profile, keep draft entry separately authorized, and review or pin the installer and dependency revisions.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (8)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk is an installation command, not an implementation of the described video-note workflow. It does not use a Bilibili URL, learning goal, or browserId, inspect visible webpage content, or generate notes. The repository name suggests a related skill, but does not establish the behavior of its contents. This assessment is limited to the supplied command.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

The installation command invokes an unpinned npm CLI through npx and references a skill repository without an immutable revision. A compromised or malicious future release could execute code with the installing user's permissions or introduce altered skill instructions. This is a supply-chain exposure, not evidence of compromise, and the finding's description of the CLI as an MCP server is inaccurate.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

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

md
Bridge and browser extension on Apple Silicon Macs and Windows, so no separate Bridge installation is needed there. Ordinary browsers and other systems should follow the current Easy WebBridge setup described in the official guide.

For a browser task the user has authorized through this Skill, check and reuse a suitable existing environment first. If a new one is genuinely needed, create one unique environment for this task with safe defaults, read back its actual environment ID, and start it. Do not ask the user to reconfirm or supply an environment name or default parameters. After confirming the correct account and `browserId`, continue the user's current task instead of stopping at setup instructions; do not ask for approval at each necessary step. Pause only for CAPTCHA/account security checks, real ambiguity about the target account or authorization, or missing local-control capability. At completion, close only the task tab group created for this run and report results and limi

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The installation command invokes the npm package skills through npx without a pinned version, allowing a newly resolved release to execute code with the installing user's privileges. This is a dependency supply-chain exposure rather than an MCP-server configuration issue; the README's installation context makes execution plausible, but there is no evidence of malicious intent or a compromised package.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

L68 permits autonomous browser-environment selection and creation despite the manifest requiring a user-specified browserId.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The final workflow permits autonomous selection or creation of a browser environment without requiring the user to supply its identity, contradicting the earlier requirement to operate only on a user-provided browserId. This ambiguity can lead an agent to infer authorization for an unintended authenticated environment and read account-specific information; the requirement to confirm the correct account and the read-only default reduce, but do not eliminate, that risk.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 7)May include surrounding context.

text
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

The renderer fixes generated note headings and explanatory text to Chinese without a language choice.

Low
Category
Output Handling
Confidence
91% confidence
Finding

The output strings hard-code Chinese for headings, learning goals, collection disclosures, and table labels; lines 40–43 likewise use Chinese labels and notices. The file provides no language opt-in or selection mechanism and documents no justification for this language constraint.

Content

No source excerpt is available for this finding.