retrospect-session

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to in direct prompt in jection because it processes data from extern al file s and use s it to direc t future write operatio n s. \n- In ges tion point s: The skill use s Glob and Read to in ges t conten t from d oc s/rules/**/*.m d and revie w s the curren t sessio n his to ry. \n- Boun dar y mar ker s: The in struc tio n s lac k delimi ter s or 'ign ore em be dde d in struc tio n s' war ni ng s whe n proce s s in g the rule file s. \n- Capabi li ty in ven to ry: The skill use s Write an d Edi t to modi fy loca l file s base d on the an al y ze d in pu t. \n- Sani ti za tio n: No sani ti za tio n or vali da tio n is de s c ri be d to pre ve n t malicio us conten t in the rule file s from bei n g exe c u t e d as agen t in struc tio n s.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 04:58 PM