qinghu-tvc-ad-film
Fail
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes several system commands including
curl,tar,npm, andsha256sumto facilitate the setup of its operating environment. It also executes theqhkitCLI tool to perform advertising film generation tasks. - [EXTERNAL_DOWNLOADS]: The skill documentation includes steps to fetch Node.js binaries from
nodejs.organd the@iqinghu/qhkitpackage from official NPM registries. These downloads target well-known and reputable service providers. - [PERSISTENCE]: The skill provides instructions for the agent to append a PATH modification command (
export PATH=...) to the user's~/.bashrcfile. While this is a common developer practice for maintaining access to binaries, it technically constitutes a persistence mechanism within the shell environment. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests untrusted data from
params.jsonfields such as '产品名称' (Product Name), '核心卖点' (Selling Points), and '目标人群' (Target Audience) as seen inSKILL.md. - Boundary markers: There are no explicit delimiters or safety instructions provided to the agent to distinguish between data and commands within these fields.
- Capability inventory: The skill has the capability to execute shell commands (
qhkit), read local product images, and perform network operations through the vendor's API. - Sanitization: The instructions do not define any sanitization or validation logic for the text strings before they are processed by the AI generation workflow.
Recommendations
- HIGH: Downloads and executes remote code from: https://nodejs.org/dist/v22.22.3/SHASUMS256.txt - DO NOT USE without thorough review
Audit Metadata