tasks-spec-update
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [In d i rect Promp t In j e c t i o n] (LOW): The skill in g e s t s data from ext e r n a l file s, crea t in g a sur f a c e for in d i rect promp t in j e c t i o n if those file s con t a in mal i c ious direc t i v e s. * In g e s t ion poin t s: The Read, Grep, an d Git (via Bash) tool s rea d con t e n t from sourc e cod e an d spec i f i c a t i o n s. * Boun d a ry mark e r s: No spec i f i c del i m i t e r s or in s t r u c t i o n s to in g n o re emb e d d e d promp t s are presen t. * Cap a b i l i t y in v e n t ory: The age n t can exec u t e shel l comma n d s (Bash) an d mod i f y file s (Writ e, Ed i t). * San i t i z a t i o n: No san i t i z a t i o n is per f orm e d on the rea d data.
- [COMMAND_EXECUTION] (SA F E): The skill use s the Bash tool for leg i t i m a t e pur p o s e s suc h as git dif f, fin d, an d grep. The comma n d pat t e r n s do n o t exh i b i t un s a f e shel l in j e c t i o n or priv i l e g e esc a l a t i o n risk s.
Audit Metadata