Back to skill

Security audit

视频换服装换场景换光线|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed AI-HIVE video-generation workflow, but it has under-scoped credential and file-write risks that users should review before installing.

Install only if you trust AI-HIVE and will keep generation commands pointed at the official service. Do not use custom --base-url or AI_HIVE_BASE_URL with a real API key, avoid running it on sensitive media without understanding that files may be uploaded, and set --no-download or a controlled output directory until the download path handling is fixed.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/videogen.py:88
Finding

Bearer API Key Disclosure Through an Unrestricted API Base URL

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/videogen.py:319
Finding

Unvalidated Server-Controlled Downloads Can Escape the Output Directory and Overwrite Files

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 (19)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a tightly scoped video-editing workflow, but observed capabilities include broader generic chat/image/model/account functions without the specialized validation it promises. That combination increases the risk of unauthorized or deceptive media generation because users may believe the skill checks consent, branding, continuity, or reference rights when it does not.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a tightly scoped video-editing workflow, but observed capabilities include broader generic chat/image/model/account functions without the specialized validation it promises. That combination increases the risk of unauthorized or deceptive media generation because users may believe the skill checks consent, branding, continuity, or reference rights when it does not.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a tightly scoped video-editing workflow, but observed capabilities include broader generic chat/image/model/account functions without the specialized validation it promises. That combination increases the risk of unauthorized or deceptive media generation because users may believe the skill checks consent, branding, continuity, or reference rights when it does not.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 81)May include surrounding context.

md
python3 scripts/videogen.py init --skill-name video-outfit-scene-light-edit-ai-hive

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 83)May include surrounding context.

md
python3 scripts/videogen.py init --skill-name video-outfit-scene-light-edit-ai-hive

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The implementation materially differs from the declared skill purpose: the manifest advertises AI-HIVE-based outfit/scene/light generative editing, but the code is only a local ffmpeg helper for trim, aspect, mute, loudness normalization, concat, and probe. This mismatch is dangerous because downstream systems or reviewers may trust the skill to enforce the advertised workflow, authorization checks, and external API handling when none of that exists, creating policy bypass, misuse, and supply-chain trust issues.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill advertises executable workflows and includes commands that rely on environment variables, file I/O, network access, and shell execution, but it declares no explicit tool scope or permission boundaries. In an agent environment, this can cause overbroad capability exposure, making it easier for the skill to access local files, invoke external services, or execute commands beyond what users expect.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation criteria include broad phrases such as generative editing and requests to turn reference cases into original content with scripts, prompts, code, or task lists. Overbroad triggers can cause the skill to activate in contexts it cannot safely or accurately handle, increasing the chance of inappropriate use for copyright-sensitive, deceptive, or policy-evasive media tasks.

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 requires a production-ready Chinese workflow, which imposes a language choice regardless of user preference. This is a natural-language policy concern because the file does not offer a language choice or justify a mandatory locale.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill enables implicit invocation (allow_implicit_invocation: true) but the metadata does not define narrow activation constraints beyond a broad description. That increases the chance the agent will trigger this skill in ambiguous contexts and automatically produce workflow/code for powerful external video-editing operations without explicit user confirmation. In a media-editing skill, mistaken activation can facilitate unauthorized image/video manipulation or policy-sensitive synthetic content generation.

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 is presented as a narrowly scoped video outfit/scene/light editing skill, but its top-level module and CLI expose a generic AI-Hive client with chat, image generation, model enumeration, upload, and account-query features. This scope expansion increases attack surface and can enable use cases outside the user-approved purpose, undermining least-privilege and policy containment.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The natural-language docstring and all user-facing CLI messages are presented only in Chinese, with no indication that the user can choose another language. This can violate language or locale policy where skills must not force a specific language without opt-in.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The standalone text-chat capability is unrelated to the declared video editing task and broadens what the skill can do beyond user expectations. Even if not invoked by the skill-specific CLI, bundling it in the same tool creates an avoidable capability expansion that may be exposed through alternate execution paths or future integration mistakes.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Standalone image generation is outside the declared scope of a video outfit/scene/light editing skill and meaningfully expands the operational surface. This violates least-functionality expectations and can let the skill be repurposed for unrelated content generation without separate review or user consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The upload flow sends local media files to a remote API/object store without any explicit confirmation, warning, or prominent disclosure at the point of use. Because this skill handles potentially sensitive videos, images, and audio, silent transmission can cause privacy and confidentiality harm if users assume processing is local or do not realize third-party storage is involved.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The get_user_info endpoint and corresponding CLI command retrieve user information and wallet balance, which are account-management capabilities rather than video outfit/scene/light editing functions. The manifest discusses media upload, configuration lookup, pricing snapshot, routing, polling, and downloads for generation tasks, but not viewing account profile or wallet data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The polling flow creates the output directory and writes downloaded files to the user's Downloads path by default. While download progress is printed during execution, the script does not clearly warn in top-level help or documentation that invoking generation will automatically save files locally unless --no-download is used.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file presents a constrained skill-specific parser while still embedding a broader generic AI toolset in the same module. This mismatch increases the risk of accidental exposure, maintenance drift, or alternative invocation paths that bypass the intended narrow interface.

Content

No source excerpt is available for this finding.