juicebox-local-dev-loop
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the Juicebox API (
/v1/search,/v1/profiles/:id). If the external API source is compromised or contains malicious content, it could influence the agent's behavior. - Ingestion points: Data is received via the
http-proxy-middlewareinsrc/dev/server.tsor directly from the Juicebox API endpoints when integration mode is active. - Boundary markers: The instructions include guidelines to "reject requests that contain credentials or contact-level data" and to start in "mock mode" with synthetic records, which serves as a safety boundary.
- Capability inventory: The skill has access to
Read,Write,Edit,Grep, andBash(npm:*)tools. - Sanitization: The skill instructions mandate logging aggregate diagnostics only and verifying suppression, though explicit code-level sanitization is not shown in the provided snippets.
Audit Metadata