The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines several methods that ingest and process external data, creating an attack surface where malicious instructions could be embedded in data processed by the agent.
Ingestion points: Untrusted data enters the context through train_model, estimate_value, calculate_market_trends, create_lease, and submit_maintenance_request in SKILL.md, and via create_listing, search_properties, and schedule_showing in references/PROPERTY_LISTING_SYSTEM.md.
Boundary markers: No specific boundary markers or "ignore embedded instructions" delimiters are present to protect the agent from interpreting data contents as instructions.
Capability inventory: The skill has access to Read, Write, and Edit tools, which could be leveraged to perform unauthorized file operations if an injection is successful.
Sanitization: The provided Python classes do not implement sanitization or validation of free-text fields like property descriptions or maintenance notes, which are common vectors for injection.

real-estate-expert