feature-blueprint
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The benchmarking script
evals/compare/run.pyutilizes thesubprocess.runfunction to executegitandclaudeCLI commands. These operations are intended to automate the skill's evaluation process by running test cases and grading the results. - [EXTERNAL_DOWNLOADS]: The evaluation suite is designed to clone an external GitHub repository (
github.com/obra/superpowers) to serve as a baseline for performance comparison during development testing. - [INDIRECT_PROMPT_INJECTION]: The skill's intake process involves reading and analyzing external materials such as project documentation, code repositories, and chat logs. This creates a surface where instructions embedded in those materials could potentially influence the agent. The skill includes specific safeguards, such as instructions to identify and exclude sensitive credentials and mandatory adherence to a structured document template.
Audit Metadata