Back to skill

Security audit

AI商品图工作室|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its stated product-image purpose, but its API client handles keys and downloads in ways that could expose credentials or overwrite local files if misdirected.

Review before installing. Use only the default AI-HIVE endpoint, avoid --base-url or AI_HIVE_BASE_URL overrides unless you fully trust the destination, keep the API key low-privilege and revocable, and download outputs only into a disposable directory until URL and filename validation are 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/imagegen.py:91
Finding

Bearer API Key Can Be Sent to an Arbitrary User-Controlled Endpoint

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/imagegen.py:312
Finding

Untrusted Task Results Permit Server-Side Request Forgery and Output-Path Traversal

Content
View full analysis
0: pct = int(downloaded * 100 / total) bar = "█" * (pct // 5) + "░" * (20 - pct // 5) print(f"\r {bar} {pct}% ({downloaded // 1024}KB)", end="", flush=True) print() size_mb = downloaded / (1024 * 1024) print(f"[ok] {out_path} ({size_mb:.1f} MB)") except requests.exceptions.RequestException as e: print(f"\n[error] Download failed: {e}", file=sys.stderr) ``` ```python task_type = task.get("taskType", "task") for i, item in enumerate(succeeded): result_url = item.get("resultUrl") if not result_url: continue ext = ".mp4" if "video" in task_type.lower() else ".png" filename = f"{task_type}_{task_id}_{i+1}{ext}" out_path = out_dir / filename download_file(result_url, out_path) # Last frame for item in succeeded: last_frame = item.get("lastFrameUrl") if last_frame: out_path = out_dir / f"{task_type}_{task_id}_lastframe.png" download_file(last_frame, out_path) ``` ### Technical Analysis The task response is treated as trusted even though it originates from a remote service. Two related validation failures are pr ...[truncated 2765 chars]
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill claims image-generation, API routing, pricing snapshot retrieval, asynchronous polling, uploads, and downloads, but the described runnable behavior does not substantiate those capabilities. This mismatch can mislead operators into trusting the skill with external calls, costs, or content-production workflows it may not safely or correctly perform, increasing the risk of unintended execution paths, unsafe substitutions, or billing surprises.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill exposes examples that invoke shell commands, environment variables, local file paths, and outbound network access, but it does not declare any corresponding tool scope or allowed-tools policy. This creates an implicit privilege boundary failure: a host may grant broader capabilities than reviewers or users expect, enabling file access, secret use, or billable API calls without explicit permission constraints.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest description and title present the skill entirely in Chinese and repeatedly frame the skill around Chinese-language invocation terms such as “AI商品图” and “产品摄影,” without offering a language choice or stating that the skill is region-specific. Under the policy, a skill should not force a specific language or locale unless the constraint is explicitly documented and justified or the user can opt in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The default prompt is broad enough to activate the skill for generic product-photo requests without requiring explicit user confirmation that AI-HIVE tooling, code generation, or external API workflows are desired. In a commerce-focused image generation skill, this can cause unintended invocation, overcollection of product/brand inputs, or accidental guidance into regulated or policy-sensitive image generation tasks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Forcing Chinese output without user opt-in can misalign the assistant response with the user's language, reducing transparency and increasing the chance the user misunderstands generated workflow steps, code, pricing, or compliance constraints. In a skill that may produce runnable commands and operational instructions, language coercion can contribute to user error and weaken informed review of sensitive actions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file is presented as a product-photo studio skill, but its implementation advertises and includes generic AI capabilities beyond that scope. Scope expansion matters because a caller expecting only product-image operations may unknowingly gain access to unrelated text and video APIs, increasing the attack surface and creating opportunities for policy bypass or unauthorized use of broader account capabilities.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

This code file contains natural-language descriptions and operational prompts entirely in Chinese, including the module docstring and later CLI help/output text. Under the language/locale policy, forcing a specific language without user opt-in is a policy violation because users are not offered any alternative locale or language selection.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Retrieving user account information and wallet balance is not necessary for generating product images, so exposing this endpoint violates least privilege and increases unnecessary data exposure. If invoked by an agent or user who only intended image generation, it could disclose sensitive account metadata or billing information without a business need.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The generic text-chat capability is outside the stated product-photo studio purpose and allows arbitrary prompt submission to a text model. In an agent setting, that creates a hidden alternate execution path that can be used to bypass intended workflow restrictions, send unrelated sensitive content to the provider, or expand the skill into a general-purpose LLM tool without review.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

Video-generation support exceeds the declared functionality of a product-photo image studio and materially broadens what the skill can do. Hidden or undocumented media-generation paths increase abuse potential, including unauthorized content creation, policy evasion, and higher-cost operations triggered through an ostensibly image-only integration.

Content

No source excerpt is available for this finding.