The Agent Skills Directory

[SAFE]: No security issues detected. The skill follows best practices for tool use and prompt engineering.
[COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/brief_builder.py, scripts/verdict_synthesizer.py, scripts/cheapest_test_designer.py) using the standard python interpreter. These scripts only perform string manipulation and arithmetic calculations based on user-provided arguments.
[PROMPT_INJECTION]: While the skill uses an 'adversarial' role-play approach (Critic, Champion, etc.), these personas are explicitly bounded by the 'Judge' persona and the deterministic tools to prevent any unintended behavior or safety bypass. The skill uses standard instructional language to set these roles.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided business ideas. It mitigates injection risks by normalizing inputs through brief_builder.py and passing the resulting brief as a contained string to subagents. The subagents have limited mandates and their outputs are synthesized by a deterministic tool.

roast