juicebox-security-basics
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes functions to process and validate external data (e.g., search queries, candidate data). These ingestion points constitute a potential attack surface for indirect prompt injection if the data contains hidden instructions. However, the skill implements defensive measures such as schema validation using the
zodlibrary.\n - Ingestion points:
validateSearchQueryfunction inSKILL.mdaccepts arbitrary data.\n - Capability inventory: The skill is configured with
Read,Write, andGreptools.\n - Sanitization: Implements schema validation via
zod.\n - Boundary markers: No explicit boundary markers or delimiters are defined for the processed input data.\n- [SAFE]: The skill demonstrates security best practices, including proper API key management (avoiding hardcoding, using environment variables), PII redaction logic for logs, and cryptographically sound webhook signature verification using timing-safe comparisons. The instructions explicitly direct the agent to use sandbox environments and verify security controls.
Audit Metadata