localization-testing
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface via external data ingestion.
- Ingestion points: The skill reads HTML body content via page.textContent('body') and processes translation files as shown in the JavaScript snippets in SKILL.md.
- Boundary markers: There are no boundary markers or instructions to ignore instructions found within the ingested content.
- Capability inventory: The skill uses browser automation (page.goto, page.fill, page.click), database access (db.users.findOne), and agent fleet coordination (FleetManager.coordinate) as documented in the implementation snippets.
- Sanitization: No sanitization or filtering of external content is specified before the content is used in logic assertions or agent tasks.
Audit Metadata