kelly-inquiry
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill connects to official Meta Graph APIs (graph.facebook.com) to send messages and utilizes environment variable names to reference authentication tokens. These operations are restricted to the skill's primary function as a communication reception desk.
- [PERSISTENCE_MECHANISMS]: The skill maintains user sessions by storing OAuth credentials in the local file system at
~/.busabase/airapps. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources including WhatsApp, Instagram, and Email.
- Ingestion points: External data enters through
scripts/ingest_inquiries.mjs(via JSON payloads) andscripts/sync_products.mjs(via CSV or JSON files). - Boundary markers: The system enforces a mandatory manual approval queue (
#/approvalsview) where users must review and approve all agent-drafted replies and quotes before any network operations occur. - Capability inventory: The skill has the capability to perform network requests via official Meta APIs and perform file writes to the Busabase database.
- Sanitization: The companion UI uses
escapeHtmlfor content rendering, and ingestion scripts perform structure validation and deduplication on incoming data payloads.
Audit Metadata