The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data from sources such as Turo bookings and guest communications, which creates an attack surface for indirect prompt injection.
Ingestion points: Turo booking ingestion and inbound communications processing as mentioned in SKILL.md.
Boundary markers: Absent. The skill instructions do not define delimiters or specific warnings to ignore instructions embedded within the ingested operational data.
Capability inventory: The skill has the ability to query and update a SQLite database located at /opt/myapp/powered_wild.db and manage the nuka_skippy_task queue for communication and API syncs.
Sanitization: Absent. No data validation or sanitization procedures are specified for the incoming external data streams.
[DATA_EXFILTRATION]: The skill identifies the specific internal network location and file path for the operational database containing sensitive business information.
Evidence: Mentions the database path /opt/myapp/powered_wild.db and the NukaSoft server IP 192.168.1.200 in the SKILL.md file.

jo