exa-security-basics
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documentation includes implementation patterns for handling user-supplied search queries passed to the Exa API.
- Ingestion points: User queries enter the system via the
safeDomainSearchandsanitizeQueryfunctions defined inSKILL.md. - Boundary markers: The skill recommends enabling Exa's native
moderation: trueparameter to filter unsafe search result content. - Capability inventory: The skill includes utility shell commands for auditing git history (
git log) and performing API connectivity checks (curl). - Sanitization: The skill provides a
sanitizeQueryfunction that utilizes regular expressions to strip HTML/template characters, removes null bytes, and enforces character length limits to mitigate injection risks.
Audit Metadata