kelly-real-estate-intel
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, such as property listings, competitor ads, and news portals, which presents a surface for indirect prompt injection attacks.
- Ingestion points: Market signals are collected from public real estate sources, listing portals, and media via the agent's browsing capabilities (documented in
SKILL.mdandREADME.md). - Boundary markers: The instructions explicitly mandate blocking unverified price claims, guaranteed appreciation, and legal advice. It also enforces a "review gate" where every signal, action, and draft must be human-approved before further processing.
- Capability inventory: The skill writes records to Busabase storage and prepares sales actions and communication drafts (WhatsApp, social media) for user review.
- Sanitization: The accompanying AirApp implementation includes HTML escaping in its UI rendering logic (e.g., in
app.jsandreal-estate-model.js) to mitigate XSS and related data presentation risks.
Audit Metadata