skill-creator
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
subprocessmodule to interface with theclaudeCLI for running evaluations and generating improved skill descriptions. The scripts use secure execution patterns, such as passing arguments as lists and providing prompts via stdin to prevent shell-based injection. - [DATA_EXPOSURE]: The skill creates a workspace for storing evaluation results, logs, and temporary command files in the
.claude/commands/directory. It manages these files within the local project scope. - [INDIRECT_PROMPT_INJECTION]: The optimization logic processes user-provided skill content and test queries. The skill mitigates risks from potentially malicious data by using XML-style delimiters in its internal prompts to maintain separation between instructions and data.
- [EXTERNAL_DOWNLOADS]: The review interface loads the well-known SheetJS library from a content delivery network (CDN) to enable spreadsheet rendering in the user's browser.
Audit Metadata