Back to skill

Security audit

3C产品结构与功能演示|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE video workflow helper with expected API-key, media-upload, and local video-processing behavior, with some scoping and dependency cautions but no hidden or destructive behavior found.

Install this only if you are comfortable using AI-HIVE for the selected product prompts and media. Confirm route, model, price, and uploaded files before running generation, keep the API key out of logs and repositories, and consider using a virtual environment with pinned dependencies.

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

Note
Location
SKILL.md:67
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:67 and scripts/videogen.py:14-18
Vulnerability Type: Unpinned third-party dependency
Risk Level: Low

Complete Code Snippets:

markdown
First run commands from this Skill directory. The script uses `https://ai-hive.iclip.cn/api` by default; it requires `requests`, while local video processing requires `ffmpeg`.
python
try:
    import requests
except ImportError:
    print("Missing dependency: requests. Run pip3 install requests", file=sys.stderr)
    sys.exit(1)

Technical Analysis

The project instructs users to install requests without specifying an audited version, lockfile, package hash, or trusted package index. As a result, installation resolves mutable package content according to the user's current Python package-index configuration.

The package name is legitimate and no malicious dependency is included in the repository. The risk arises from the absence of dependency-integrity controls: a compromised upstream release, compromised package index, malicious index configured with precedence, or dependency substitution in the installation environment could cause unintended code to run during package installation or import.

Attack Path

  1. A user runs the documented installation command, pip3 install requests.
  2. pip resolves the dependency through the package indexes configured in the user's environment.
  3. An attacker compromises an applicable package source, release, mirror, or package-resolution configuration.
  4. The installer retrieves the attacker-controlled package content because no version or hash constraint detects the substitution.
  5. Malicious installation or import-time code executes with the privileges of the user running the command.
  6. The malicious package may access files, environment variables, API credentials, and network resources available to that user.

Impact Assessment

Successful ...[truncated 554 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin requests and its transitive dependencies to reviewed versions in a requirements or lock file.
  2. Generate and enforce cryptographic hashes, for example by installing with pip install --require-hashes -r requirements.txt.
  3. Document an approved HTTPS package index and prevent unintended fallback to untrusted indexes.
  4. Add automated dependency vulnerability and integrity scanning to the release process.
  5. Periodically update pinned versions through a controlled review process rather than resolving the latest release during installation.
  6. Run dependency installation and the Skill under a non-privileged account or isolated virtual environment.
Vulnerability Patterns
  • 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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (9)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The declared description promises a full 3C product demo skill with visualization-oriented decomposition of product functions and optional AI-HIVE-based material generation. The supplied code does not implement those capabilities. It simply collects a few inputs and emits a local JSON draft labeled as a production blueprint. This is a materially narrower and different primary purpose than the declared end-to-end content generation and API orchestration workflow, so the description does not accurately represent the actual behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description promises a domain-specific 3C product demo creation skill with asset ingestion, structured content generation, and optional AI-HIVE-based media generation workflow. The supplied code instead implements only a generic local video-editing utility around ffmpeg/ffprobe. While video trimming/aspect conversion could support a broader media workflow, they do not substantively match the claimed primary purpose or advertised capabilities. There are no undeclared dangerous actions, but there is a clear description-to-behavior mismatch because the implementation is far narrower and materially different from the declared skill behavior.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill advertises and demonstrates code paths that can use environment variables, local files, shell commands, and network access, but it does not declare any explicit tool scope or permissions boundary. In an agent environment, this increases the chance of overbroad execution, accidental secret exposure, unintended file access, or network calls beyond what users and reviewers expect.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill enables allow_implicit_invocation without any visible trigger constraints, so an agent may auto-select this skill based on broad similarity rather than clear user consent. Because this skill can generate runnable AI-HIVE commands and production workflows involving uploads, pricing queries, routing choices, polling, and downloads, unintended invocation could cause external actions, unnecessary cost, or leakage of sensitive product assets.

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
93% confidence
Finding

The file-level docstring and the generic CLI expose broad chat, image generation, model listing, user-info, upload, and arbitrary video-generation capabilities as a '通用 AI 能力调用工具'. In contrast, the manifest describes a specialized skill for 3C product structure/function explainer content; the code therefore implements and documents substantially broader behavior than that declared scope.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring and all user-facing CLI guidance are written in Chinese, presenting the tool as a general-purpose skill rather than a region-specific one. Under the language/locale policy, this effectively forces a specific language without opt-in or an explicit justification that the skill is Chinese-only.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Although the skill is presented as a narrowly scoped 3C video-explainer tool, the file still contains and exposes generic text-chat and image-generation command paths through the broader CLI entrypoints. In a skill ecosystem that relies on metadata-based trust or policy gating, this creates a capability mismatch that can let callers use undeclared model operations, bypass least-privilege expectations, and process arbitrary user-supplied content beyond the reviewed scope.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

This code embeds the skill name and description entirely in Chinese in the module docstring and argument parser description. That can amount to a language-policy issue because the skill forces a specific language presentation without any user opt-in or alternative locale handling.

Content

No source excerpt is available for this finding.