T09 · Insecure Skill Coding Practices
- Location
scripts/videogen.py:395- Finding
Server-Controlled Path Traversal in Generated File Downloads
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill largely does AliExpress video generation as advertised, but it needs review because unsafe endpoint and download handling could expose the API key or overwrite local files.
Review before installing. Use a dedicated, revocable AI Hive API key, avoid custom AI_HIVE_BASE_URL or --base-url values unless you fully trust the endpoint, run it in an isolated environment, and keep downloads pointed at a non-sensitive output directory.
scripts/videogen.py:395Server-Controlled Path Traversal in Generated File Downloads
scripts/videogen.py:91Bearer API Key Can Be Disclosed to an Arbitrary Configured Endpoint
SKILL.md:68Unpinned and Unverified Runtime Dependency Installation
The skill includes executable shell commands and invokes a Python script that can use environment access, file I/O, network, and shell capabilities, but the manifest does not declare any tool scope or permissions boundaries. This creates an unsafe trust gap: an agent or reviewer cannot tell what capabilities are intended, and the skill could be run with broader privileges than necessary.
Because this is a markdown file, vague-trigger review applies. The description says "Use this skill for" followed by many generic phrases such as product demos, installation tutorials, ad assets, and multilingual versions, but it does not define explicit invocation phrases, scope limits, or negative examples; this breadth could overlap with many ordinary ecommerce video-editing requests and cause unintended activation.
The prompt explicitly instructs the model to "Create a Spanish-market visual base," which is a locale-specific directive. The file does not indicate that Spanish is optional, user-selected, or required for a documented region-specific compliance purpose, so this can violate the language/locale choice policy.
The file is presented as an AliExpress ecommerce video skill, but it embeds a broad AI Hive client supporting chat, image generation, model discovery, uploads, and account-related operations. This scope expansion violates least privilege and increases the attack surface, making it easier for downstream agents or users to invoke capabilities unrelated to the declared purpose and potentially route sensitive data to unintended endpoints.
This Python skill presents its top-level description, CLI guidance, status messages, and embedded skill metadata in Chinese, including the title and usage-facing strings. Because the file does not offer a language/locale choice or state that the skill is intentionally limited to a Chinese-speaking or region-specific context, it violates the language/locale policy for natural-language behavior.
The skill exposes a text chat command even though its declared purpose is AliExpress video generation/editing. In an agent setting, this creates covert capability expansion: prompts, secrets, or business data intended for video workflows could be redirected into a general chat API, enabling unintended data processing and policy bypass.
The standalone image generation command exceeds the declared scope of a video-focused ecommerce skill. This broadens what an invoking agent can do with the same credentials and infrastructure, increasing risk of unauthorized content creation, unexpected billing, and misuse of uploaded reference media outside the intended workflow.
This markdown file contains user-facing natural-language content exclusively in Chinese across the substantive changelog entries. Under the policy, forcing a specific language without user opt-in or a documented justification is a natural-language policy violation.
The string literal "请替换为你的APIKey" requires Chinese to understand the placeholder guidance. This is a natural-language locale policy issue because it forces a specific language without any opt-in or alternative.
Querying user information and wallet balance is not necessary for generating or editing listing videos. In a delegated-agent environment, this exposes account metadata and spending information that can aid reconnaissance, leak sensitive operational details, or be used for abuse planning.
Model enumeration allows callers to discover all AI Hive models instead of only the fixed AliExpress video models needed by the skill. This enables capability reconnaissance and makes it easier to pivot into unsupported or higher-risk model workflows using the same environment and credentials.