skill-creator
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill automates the development and testing workflow by executing the
claudeCLI and supporting Python scripts (e.g.,run_eval.py,improve_description.py) usingsubprocess.runandsubprocess.Popen. These actions are managed within a development context to iterate on skill quality. - [EXTERNAL_DOWNLOADS]: The evaluation viewer (
viewer.html) references the SheetJS library fromcdn.sheetjs.com. SheetJS is a well-known service, and this reference is used legitimately to enable spreadsheet rendering within the review interface. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and skill content as part of its optimization loop. While this creates a potential surface for indirect injection, the tool uses structured prompting and XML-style tags to isolate content and reduce the risk of instruction leakage.
- [DYNAMIC_EXECUTION]: The skill facilitates the creation and execution of specialized scripts (such as
create_docx.py) within a new skill's ecosystem to automate deterministic tasks, which is an intended feature of the authoring process.
Audit Metadata