Back to skill

Security audit

AI模型任务失败修复|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is not clearly malicious, but it combines repair guidance with billable AI generation, credential storage, uploads, downloads, and unvalidated API endpoints that need review before use.

Review this skill before installing. Use it only with an AI-HIVE API key you are willing to expose to this tool, avoid custom base URLs unless you fully control them, do not let agents run generation commands implicitly, confirm costs manually before image/video submissions, and only upload files you intend to send to AI-HIVE.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (4)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/imagegen.py:91
Finding

API Credentials Can Be Transmitted to an Arbitrary Server

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/imagegen.py:275
Finding

Unvalidated Server-Supplied URLs Permit Arbitrary Upload and Download Requests

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
scripts/imagegen.py:18
Finding

Unpinned Runtime Dependency Installation Creates Supply-Chain Risk

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/imagegen.py:470
Finding

Paid Generation Tasks Are Submitted Without Enforced Cost Confirmation

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (27)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Instead of focusing on diagnosing failed tasks and planning safe retries, the skill documentation emphasizes submitting new generation jobs, querying account state, storing credentials, and downloading outputs. In a repair context this is especially dangerous because retries and fresh submissions can cause duplicate billing, mishandled secrets, or data leakage if users believe the skill is only analyzing prior failures.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Instead of focusing on diagnosing failed tasks and planning safe retries, the skill documentation emphasizes submitting new generation jobs, querying account state, storing credentials, and downloading outputs. In a repair context this is especially dangerous because retries and fresh submissions can cause duplicate billing, mishandled secrets, or data leakage if users believe the skill is only analyzing prior failures.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Instead of focusing on diagnosing failed tasks and planning safe retries, the skill documentation emphasizes submitting new generation jobs, querying account state, storing credentials, and downloading outputs. In a repair context this is especially dangerous because retries and fresh submissions can cause duplicate billing, mishandled secrets, or data leakage if users believe the skill is only analyzing prior failures.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Instead of focusing on diagnosing failed tasks and planning safe retries, the skill documentation emphasizes submitting new generation jobs, querying account state, storing credentials, and downloading outputs. In a repair context this is especially dangerous because retries and fresh submissions can cause duplicate billing, mishandled secrets, or data leakage if users believe the skill is only analyzing prior failures.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file implements local video editing operations that materially diverge from the declared AI task-repair purpose. Capability mismatch is dangerous because it expands the skill's attack surface and can enable unauthorized media manipulation or hidden processing behavior outside expected repair workflows.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The file is implemented as a broad AI-Hive client for chat, model enumeration, media upload, image generation, video generation, task polling, and account utilities, which materially exceeds the declared skill purpose of task-failure repair. In an agent environment, this capability expansion increases the attack surface and enables unintended content generation or account actions under the guise of a narrowly scoped repair skill.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill advertises executable workflows that use environment variables, local files, network access, and shell commands, but it declares no explicit tool scope or permissions. That creates an authority ambiguity where the runtime may expose broader capabilities than reviewers or users expect, increasing the chance of unintended file access, network exfiltration, or command execution.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

The skill name and all user-facing instructions are presented only in Chinese, with no indication that the user may choose another language or locale. This can constitute a language/locale policy issue when the skill appears to assume Chinese as the required interaction language without explicit opt-in or justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The activation list includes generic phrases such as '任务修复' and a broad condition about wanting to turn a reference case into original content with scripts, prompts, code, or a task list. These phrases can overlap with many ordinary requests outside this skill's intended AI-HIVE task-repair scope, increasing the chance of unintended invocation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The usage guidance expands from fixing failed tasks into transforming reference cases into original marketing content, which is a materially different function. That scope drift makes the skill harder to review and can be abused to funnel users from troubleshooting into content generation workflows involving copyright, brand, and billing risks.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

A skill framed as task-failure repair also instructs users to create new commercial images and videos, broadening it from remediation into production content generation. This increases the chance of accidental billable actions, policy bypass through misleading categorization, and misuse of provided assets beyond the user's original troubleshooting intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest text says the skill will turn requests into a "production-ready Chinese workflow," which imposes a specific language choice by default. This is a natural-language policy concern because the file does not indicate any user opt-in or language selection mechanism.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest describes a skill for diagnosing failed AI model tasks, deciding retries, performing model fallback, and handling AI-HIVE API workflows such as polling and downloads. This file instead only collects project metadata and writes a local JSON 'production brief' blueprint, with no failure analysis, retry logic, API interaction, or task-state recovery behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script hard-codes Chinese-language skill naming and defaults, such as the docstring and argparse description using Chinese text, without indicating that users may choose another language or locale. This can violate language/locale policy when a skill implicitly enforces one language without explicit opt-in or justification.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The code introduces subprocess-driven ffmpeg media processing unrelated to the stated task-repair function, increasing operational and security risk. Even without shell injection, invoking powerful local binaries on user-controlled files can expose the environment to malformed-media parser bugs, resource exhaustion, and unreviewed file-processing behavior.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/edit_video.py (reported line 20)May include surrounding context.

python
def run(args: list[str]) -> None:
    print(" ".join(args))
    subprocess.run(args, check=True)


def main() -> None:

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest frames this skill as a repair/remediation tool for failed AI-HIVE tasks: diagnosing auth/parameter/material/model/async-state issues and proposing retry or recovery actions. However, the code is a general-purpose AI Hive client that directly invokes text chat, image generation, video generation, media upload, polling, and download workflows, which goes well beyond failure-repair behavior and turns the skill into a content-generation tool.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The top-level docstring says this file is a '通用 AI 能力调用工具' and lists text chat, image generation, video generation, model query, media upload, and task polling. Later fixed skill metadata identifies the skill specifically as 'AI模型任务失败修复', so the inline documentation presents a materially different intent than the packaged skill's declared purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This codebase presents its primary natural-language interface entirely in Chinese, including the module docstring and all user-facing CLI guidance, without offering a language choice or documenting that the skill is Chinese-only. That can violate language/locale policy where user-facing skills must not force a specific language without opt-in.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The module docstring and CLI framing present the code as a general-purpose AI capability tool rather than a repair-specific utility, creating a mismatch between declared intent and actual behavior. This kind of misrepresentation can cause an orchestrating agent or reviewer to grant broader trust than warranted, leading to unsafe invocation paths and policy bypass through scope confusion.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file contains natural-language descriptions and CLI guidance entirely in Chinese, including the module docstring and later user prompts/help text. Under the stated policy, forcing a specific language or locale without offering a choice is a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The script's visible user interface strings, including dependency errors, API-key setup instructions, upload/download notices, task polling updates, and argparse help text, are consistently presented only in Chinese. This effectively enforces a locale on all users without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The docstring says the script will create a deterministic production brief for 'AI模型任务失败修复|AI-HIVE', implying output directly related to repair workflows. In practice, the generated JSON contains marketing/project planning fields like audience, goal, platform, format, and required facts, not repair plans, failure classifications, retry decisions, or recovery actions.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

This code creates parent directories and writes a JSON file to the path provided by --output, but there is no confirmation prompt, warning comment, or user-facing disclosure before performing the write. Because file modification is a safety-relevant operation in code files, it should be clearly disclosed unless already made explicit to the user.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest focuses on receiving failed responses, model configuration, routing strategy, retry history, and generating recovery plans or retry actions. The code also exposes a user-info endpoint and CLI command advertised as querying user information and wallet balance, which is not part of task failure repair semantics.

Content

No source excerpt is available for this finding.