style-transform
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWNO_CODE
Full Analysis
- [NO_CODE] (SAFE): The provided skill consists entirely of JSON data files (
config/andtemplates/). No Python, JavaScript, or shell scripts are present, eliminating risks associated with malicious code execution. - [PROMPT_INJECTION] (SAFE): Analysis of the text templates and style rules found no evidence of instructions designed to bypass agent safety filters or override system prompts.
- [DATA_EXFILTRATION] (SAFE): There are no hardcoded secrets, API keys, or sensitive local file paths. No network-capable commands (e.g., curl, fetch) are defined.
- [REMOTE_CODE_EXECUTION] (SAFE): The skill does not define any external dependencies or download mechanisms for remote content.
- [INDIRECT_PROMPT_INJECTION] (LOW): The
podcast_templates.jsonfile uses placeholders like【主题】and【自定义主题】for user-provided data. While this creates a surface for indirect prompt injection, the skill's lack of executable capabilities (it only generates text) limits the impact to the generated content itself.
Audit Metadata