Back to skill

Security audit

数字人手机展示视频|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This is a plausible AI-HIVE video-generation helper, but it needs review because it handles API keys and network media transfers with weak destination controls.

Install only if you trust the publisher and AI-HIVE workflow, use a limited API key, avoid custom base URLs unless you control the endpoint, and review every generated command before running it. Keep uploads limited to media you are allowed to share, and monitor the configured ~/.ai-hive/config.json value because it controls where future authenticated requests go.

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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/videogen.py:64
Finding

Bearer API Credential Can Be Transmitted to an Arbitrary Configurable Server

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/videogen.py:238
Finding

Server-Supplied Upload and Download URLs Are Used Without Destination Validation

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
SKILL.md:86
Finding

Third-Party Python Dependency Is Installed Without Version or Integrity Pinning

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared description presents a much broader production skill centered on creating digital-human mobile showcase videos and, when needed, orchestrating AI-HIVE OpenAPI operations to generate materials and deliver outputs. The supplied code does not perform those actions. It only parses command-line arguments and writes a local JSON file with preset fields. While the blueprint concept is loosely related to pre-production planning, the actual behavior is far narrower than the declared purpose and omits the core advertised capabilities. Therefore this is a material description-behavior mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

This is a strong mismatch. The declared description describes an end-to-end specialized content-production skill for digital-human phone demo ads, with domain-specific inputs, outputs, and remote API orchestration. The supplied code does none of that. It is a narrowly scoped, auditable CLI wrapper around ffmpeg/ffprobe for deterministic editing operations on existing videos. While generic video editing could be a supporting implementation detail within a larger media pipeline, this code chunk alone lacks the core advertised behavior and instead has a materially different primary purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill includes executable examples and references to capabilities such as environment variables, file access, shell commands, ffmpeg, and network/API usage, but it does not declare any explicit tool scope or permission boundaries. In an agent environment, this creates an over-privileged or ambiguously privileged skill surface where the runtime may allow sensitive operations without clear review, increasing the risk of secret exposure, arbitrary file modification, or unintended external requests.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The trigger phrases include broad terms like app advertising, product showcase, and phone screen video, which can cause the skill to activate for many generic marketing requests beyond its intended niche. Overbroad invocation expands the attack surface by increasing the chances that the agent routes unrelated user tasks into a skill that can suggest networked, file, or shell-based workflows, potentially leading to confused-deputy behavior or unnecessary exposure of sensitive inputs.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The invocation text is broad enough that the skill may be selected for generic requests about turning a request into a workflow, even when the user did not ask for avatar-phone showcase content. In this skill, over-invocation matters because it can steer users into a specialized marketing/video pipeline, potentially causing inappropriate tool use, unintended API actions, or irrelevant outputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The default prompt explicitly forces Chinese output regardless of the user's language preference. This can degrade user comprehension, conceal important safety or consent details from non-Chinese-speaking users, and increase the chance of mistaken approval of generated workflows or API actions.

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
96% confidence
Finding

The file presents itself as a phone-showcase video skill, but the implementation and docstring expose a broad, reusable AI utility for chat, image, video, model enumeration, upload, and task polling. In an agent-skill setting, this capability mismatch expands the reachable attack surface and may let callers invoke functions outside the reviewed business scope, undermining least privilege and policy controls.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The top-level documentation describes a generic AI invocation tool, while later code frames the file as a specialized avatar phone showcase skill. This inconsistency is dangerous in security review and deployment because operators may grant trust and permissions based on the narrow skill identity while the file actually contains a much broader toolset.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This code file contains user-facing natural-language descriptions entirely in Chinese, including the module docstring and CLI help text, without offering a language choice. Under the policy rule, forcing a specific language or locale without explicit user opt-in is a natural-language policy violation unless the constraint is clearly justified, which is not documented here.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The chat endpoint gives the skill a generic text-generation capability unrelated to avatar phone showcase video creation. In an agent environment, this can be repurposed for arbitrary prompt execution, policy bypass attempts, or unreviewed content generation beyond the intended workflow, making the skill materially more dangerous than advertised.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Standalone image generation is outside the stated video-focused purpose and enables broader synthetic-content creation than users or reviewers would expect from this skill. That mismatch weakens trust boundaries and could be abused to generate unrelated assets or deceptive materials under the cover of a specialized marketing-video tool.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The script hard-codes Chinese-language natural-language identifiers and descriptions such as the module docstring and argparse description. This imposes a specific language/locale on users without offering opt-in or documenting that the skill is intentionally region-specific, which matches the locale policy violation criteria.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The user-info command exposes account metadata and wallet/balance information that is unrelated to generating showcase videos. While not code execution, it creates unnecessary access to sensitive account details and increases the chance of data exposure or misuse through an otherwise specialized skill.

Content

No source excerpt is available for this finding.