ClawHub
Back to skill

Security audit

汽车/赛车宣传短片｜AI-HIVE原创工作流

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent AI-HIVE advertising workflow with disclosed remote MCP use and no evidence of hidden exfiltration, destructive behavior, or persistence.

Install only if you intend to use AI-HIVE for advertising image or video workflows. Prefer OAuth or a client secret store, keep API keys out of prompts/logs/screenshots, use the documented https://ai-hive.iclip.cn/api/mcp endpoint, and confirm cost-bearing generation or publishing actions separately.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (4)

Tainted flow: 'request' from os.environ.get (line 67, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Content
method="POST",
    )
    try:
        with urllib.request.urlopen(request, timeout=60) as response:
            result = parse_payload(response.read(), response.headers.get("content-type", ""))
            return result, response.headers.get("mcp-session-id") or session_id
    except urllib.error.HTTPError as error:
Confidence
83% confidence
Finding
The script reads API credentials from environment variables and sends them to the remote MCP endpoint selected by AI_HIVE_MCP_URL. If that environment variable is overridden to an attacker-controlled host, the helper will transmit the API key or bearer token in request headers, enabling credential theft and arbitrary remote tool invocation.

Lp3

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding
The skill instructs users to run local scripts, export an API key, connect to a remote MCP endpoint, and write output files, which effectively grants network, environment-variable, file-read, and file-write capabilities without an explicit permission declaration. Even though the workflow includes sensible cautions, the undeclared capability surface increases the risk that a hosting platform or user underestimates what the skill can access or cause a connected agent to do.

Description-Behavior Mismatch

Medium
Confidence
88% confidence
Finding
The helper exposes generic remote tools/list and tools/call functionality rather than constraining operations to the advertised automotive advertising workflow. In a skill context, this broad capability increases the blast radius if the remote MCP server offers unrelated or sensitive tools, and users may invoke them with the supplied credentials under the assumption the skill is purpose-limited.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The manifest enables implicit invocation for a skill that can access a remote MCP service capable of generating marketing, image, and video outputs. Because the trigger conditions are only described loosely in metadata and not enforced in the policy itself, the skill may be auto-invoked in broader situations than the user intended, causing unintended tool use, remote data disclosure to the MCP endpoint, or unexpected content-generation flows.

Static analysis

No suspicious patterns detected.