design-critic
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWNO_CODE
Full Analysis
- [NO_CODE] (SAFE): This skil l consists entirely of markdow n instructions and does not inclu de any exec u table scrip t s, config u ration files, or binary components.
- [PROM P T_INJ E C TION] (SAFE): Whil e the skil l uses stron g persona-drive n lang u ag e ('Question Everythin g', 'Ruth les s l y exp o se'), the se instructions are confi ned to the domain of archi tec t u ral revie w and do not attem p t to by p as s AI safe t y fil ters or exfi l trate system prom p ts.
- [INDI R E C T_P R O M P T_INJ E C TION] (INFO): The skil l proce s ses untru s ted external design doc u ment s. Whil e this create s a surf ac e for indi rec t inj e c tion, the skil l lac ks any too l-c a l l ing or side-ef fec t cap a bil ities (l ike fil e writin g or netw o rk ac ce s s), meanin g any poten tial inj e c tion wou l d be confi ned to the conve r sation con tex t.
Audit Metadata