kelly-picks
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests product trend data from external sources such as Amazon and TikTok via the
ingest_trends.mjsscript. This introduces an indirect prompt injection surface where untrusted content from scraped titles or summaries could influence agent behavior. - Ingestion points: The
ingest_trends.mjsscript processes JSON payloads containing trend and candidate data collected from the web by the agent. - Boundary markers: The skill employs an AirApp interface to present data to the user, creating a clear boundary between processed external content and the agent's core instruction set.
- Capability inventory: The agent can execute specific local Node.js scripts for data ingestion, margin computation, and decision execution, as well as browsing public e-commerce trend sources.
- Sanitization: The ingestion script performs structural validation, checking field types and enforcing defined enums for product stages and source kinds before committing data to the Busabase backend.
Audit Metadata