Security audit

bilibili-search

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small Bilibili search wrapper that forwards a user-provided keyword to a local search service, with no evidence of deception, persistence, or destructive behavior.

Install only if you intentionally run and trust the companion local FastAPI service on 127.0.0.1:8000. Avoid putting sensitive information in search keywords, since the backend may browse Bilibili or log requests depending on that local service's implementation.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill explicitly advertises headless background browser automation and real-time web retrieval, but the documentation does not warn users that it will initiate network access and browse websites on the local machine. In an agent setting, undocumented background browsing can create privacy, consent, and monitoring risks because users may not realize the skill triggers outbound requests and local browser automation.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The tool sends user-supplied search terms to a local HTTP service automatically and is explicitly configured with require_approval: "never", so users are not warned that their input is being transmitted off-tool. Even though the destination is localhost, local services can expose sensitive data, trigger side effects, or bridge to other internal resources; the lack of transparency and consent makes this a real privacy and trust-boundary issue.

Natural-Language Policy Violations

Low

Confidence: 84% confidence
Finding: The trigger examples and user instructions are presented only in Chinese, and the skill is framed around interacting with Bilibili in that language context. There is no indication that users may choose another language or that the Chinese-only interaction is a justified locale-specific limitation.

Natural-Language Policy Violations

Low

Confidence: 86% confidence
Finding: The natural-language description forces a specific language locale in the tool metadata without offering the user a language choice or documenting a justified locale restriction. This can violate language or locale policy when the skill is expected to be generally usable across users.

Static analysis

No suspicious patterns detected.