Back to skill

Security audit

广告疲劳检测与AI焕新|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to do what it says: help create AI-HIVE ad-refresh workflows, while requiring users to manage API keys, selected media uploads, and possible paid generation tasks carefully.

Before installing, be comfortable with AI-HIVE network use, storing an API key locally, uploading only media you are authorized to use, and reviewing final parameters before any generation that may incur cost. Keep API keys out of prompts, logs, screenshots, and version control.

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
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
97% confidence
Finding

The skill is presented as a narrowly scoped ad-optimization tool, but the underlying behavior reportedly includes broader AI-HIVE functions such as user info lookup, wallet/balance queries, model enumeration, and generic generation. That kind of undeclared expansion materially increases risk because a user may invoke a seemingly harmless marketing skill that can access account-level or billing-adjacent functions beyond the advertised purpose.

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 narrowly scoped ad-optimization tool, but the underlying behavior reportedly includes broader AI-HIVE functions such as user info lookup, wallet/balance queries, model enumeration, and generic generation. That kind of undeclared expansion materially increases risk because a user may invoke a seemingly harmless marketing skill that can access account-level or billing-adjacent functions beyond the advertised purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill is presented as a narrowly scoped ad-optimization tool, but the underlying behavior reportedly includes broader AI-HIVE functions such as user info lookup, wallet/balance queries, model enumeration, and generic generation. That kind of undeclared expansion materially increases risk because a user may invoke a seemingly harmless marketing skill that can access account-level or billing-adjacent functions beyond the advertised purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill advertises executable examples that rely on environment variables, local file access, shell execution, and network calls, but it does not declare any explicit tool scope or permissions boundary. In an agent environment, this mismatch can cause over-privileged execution, unclear operator expectations, and accidental exposure of local files or secrets if the runtime grants broad defaults.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The trigger phrases include broad generic terms like advertising optimization and information-flow creatives, which may cause the skill to activate in contexts where the user did not intend to use this capability. Over-broad activation increases the chance of unnecessary data collection, confusing tool selection, or accidental execution of networked/media-processing steps.

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 hardcodes Chinese output regardless of the user's language preference, which can override user intent and reduce transparency about what the system is doing. In a workflow that produces operational instructions and runnable commands, forcing a language without opt-in may also impair review, oversight, or safety checks by users who expect another language, making misuse or mistakes harder to detect.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest enables implicit invocation, but the activation conditions are only broadly described in metadata and do not define clear boundaries or exclusions. Because this skill can generate runnable workflows and API-driving commands in a sensitive advertising domain, automatic triggering could cause the agent to invoke ad-generation or optimization behavior in contexts the user did not explicitly authorize, increasing the risk of unintended policy-bypassing assistance or unsafe marketing output.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This Python file contains user-facing natural-language strings entirely in Chinese for the skill name, description, and search aliases, with no indication that the skill is region-specific or that users can choose another language. That creates a locale policy issue because the skill imposes a specific language context by default rather than offering choice or documenting a justified constraint.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest describes a skill focused on ad-fatigue detection, retention/replacement variable analysis, and AI-assisted refresh generation through AI-HIVE workflows. Line L26 declares output artifacts including “差异化复刻方案” (replication plan), which expands the skill into producing replication-oriented deliverables not clearly justified by the stated purpose and is in tension with the manifest’s warning against unauthorized copying.

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:

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest describes a skill focused on diagnosing ad fatigue from performance data and, when generating refreshed creatives, interacting with AI-HIVE OpenAPI for configuration lookup, uploads, routing, polling, and downloads. This file instead exposes generic local video-editing operations such as trim, aspect-ratio conversion, muting, loudness normalization, probing, and concatenation via ffmpeg, which are not mentioned as part of the skill’s declared purpose or integration model.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest describes a domain-specific workflow for advertising fatigue detection and creative refresh based on historical performance data, preserved/replaced variables, and refresh batches. In this file, the implemented skill entrypoints at L832-L879 only generate/query/upload media through AI-Hive and do not perform fatigue analysis, CTR/CVR-based diagnosis, variable-retention logic, or refresh-batch planning.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The module docstring presents the skill entirely in Chinese as a '通用 AI 能力调用工具' and the surrounding user-facing strings throughout the file are also Chinese-only. For a general-purpose skill, this indicates a fixed language/locale choice without any user opt-in or documented regional constraint, which matches the language-policy violation criteria.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifested skill is framed as a focused ad-fatigue refresh capability, but this file embeds broad reusable commands for user-info lookup, model enumeration, text chat, image generation, and arbitrary video generation. Those capabilities are more general than the declared purpose and are not justified by the specialized advertising-diagnosis workflow described in the manifest.

Content

No source excerpt is available for this finding.