chrome-extension-mv3
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill documen t s patterns where content scrip t s inges t untrus t ed data from the DOM, creat i n g a surface for indirec t prom p t injec t i o n.
- Inges t i o n point s: DOM tex t extrac t i o n (documen t.body.innerTex t) described in refer e n c e s/content -scrip t s.md.
- Boundary marker s: None ident ified; instruc t i o n s do not sugges t wrap p i n g extern al data in delimiter s.
- Capabilit y invent ory: Patterns include chrome.runtime.sendMessage and chrome.scrip t ing.executeScrip t in refer e n c e s/content -scrip t s.md and refer e n c e s/message -pass i n g.md.
- Sanit i z at i o n: No specific sanit i z at i o n or validat i o n logic is sugges t ed for proces s ed DOM content.
- [EXTERNAL_DOWNLOADS]: The skill refer e n c e s an extern al ut ilit y for searc h i n g develo pm e n t guideline s.
- Sourc e: Recommend s the modern -web -guidance package from the GoogleChrome organi z at i o n.
- Mechanis m: Instruc t s the agent to use npx -y modern -web -guidance@lates t to searc h for document at i o n.
Audit Metadata