Back to skill

Security audit

Seedream 5.0 Lite 营销图片

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly an AI Hive marketing-image generator, but it handles API credentials in an under-scoped way that can send the user's key to an arbitrary configured endpoint.

Review before installing. Use this only with a trusted AI Hive API key, avoid setting --base-url, AI_HIVE_BASE_URL, or config base_url to anything except the intended AI Hive HTTPS endpoint, and install dependencies in an isolated environment. Be aware that init stores an API key under ~/.ai-hive/config.json and that reference images you choose are uploaded to the provider.

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 (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:71
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis
Remediation
View remediation
" ``` 2. Maintain dependencies in a locked requirements file with cryptographic hashes: ```text requests== --hash=sha256: ``` 3. Install with hash enforcement: ```bash python3 -m pip install --require-hashes -r requirements.txt ``` 4. Explicitly use a trusted package index and avoid untrusted global pip configuration. 5. Recommend installation in a dedicated virtual environment with only the permissions required by the Skill. 6. Periodically review and deliberately update the pinned version after security testing. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/imagegen.py:90
Finding

API Key Disclosure Through an Unrestricted Custom Base URL

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (10)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This markdown file contains user-facing natural-language content entirely in Chinese across the substantive changelog entries. Per the policy, forcing a specific language without user opt-in is a natural-language policy violation unless a justified locale constraint is documented, which is not present here.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill invokes shell commands, installs dependencies, performs networked image-generation tasks, and reads/writes files, but it does not declare any explicit tool scope or permissions. This creates a least-privilege failure: an agent may execute broader capabilities than a reviewer or runtime policy expects, increasing the chance of unintended command execution, filesystem access, or outbound network use.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description and main instructions are written as Chinese-only operational guidance, while also embedding an English invocation hint, which indicates mixed-audience use. There is no statement that the skill supports multiple languages or that Chinese is required for a region-specific compliance reason, so this appears to impose a locale preference without user opt-in.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The file is packaged as a narrowly scoped Seedream 5 Lite marketing-image skill, but the implementation is a full generic AI Hive client with broader capabilities. This violates least-privilege and increases attack surface: a caller or wrapper that invokes the script outside the intended happy path can access unrelated operations not justified by the advertised skill.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The module docstring explicitly presents the file as a generic AI-capability tool, contradicting the narrow marketed role of the skill. This mismatch is dangerous because it signals capability overreach and can mislead reviewers, orchestration systems, or users about the true behavior and privileges of the artifact.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This code file embeds user-facing descriptions, prompts, and operational guidance in Chinese, starting with the module docstring and continuing throughout the CLI UX. Because the skill does not offer any language selection or note that it is intentionally region-specific, it effectively imposes a locale on all users.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Text-chat capability is unrelated to a fixed marketing-image skill and enables arbitrary prompt transmission to a general text model path. In a skill ecosystem, this can be abused as capability smuggling: consumers may believe they are granting only image-generation behavior while the code retains broader data-processing functionality.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

Video-generation support materially exceeds the declared scope of a Seedream Lite marketing-image skill. Extra media-upload and generation paths expand the opportunity for misuse, policy bypass, and unexpected resource consumption under the guise of a simpler image-only skill.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

User-info and wallet-balance retrieval are not necessary to generate marketing images and expose account metadata beyond the skill’s stated purpose. While not code execution, this is an unnecessary data-access capability that broadens privacy and reconnaissance risk if the skill is trusted more than its real behavior warrants.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

Enumerating arbitrary models undermines the claim that this is a fixed Seedream 5 Lite skill and provides discovery of broader platform capabilities. This can facilitate misuse by revealing alternative models and routes, enabling callers to pivot beyond the expected function of the skill.

Content

No source excerpt is available for this finding.