immoscout24-landlord-outreach
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external web pages (ImmoScout24 listings) to identify provider names and listing details. This creates a surface for indirect prompt injection if listing text contains adversarial instructions.
- Ingestion points: The agent reads content from listing expose URLs, including provider blocks and listing descriptions as described in
SKILL.md. - Boundary markers: No explicit delimiters or boundary markers are defined for the data ingested from the browser.
- Capability inventory: The skill utilizes browser-internal automation to clear textareas, input personal data, and submit contact forms (
SKILL.md). - Sanitization: The skill relies on basic keyword verification (e.g., checking for 'Tauschwohnung') but does not specify technical sanitization of the listing text.
- [SAFE]: The skill follows secure context management by instructing the agent to load personal facts from a dedicated source (
shared/personal-profile.yaml) rather than using hardcoded values. It also emphasizes submission safety, requiring explicit user approval before transmitting personal information to external providers.
Audit Metadata