exa-search
Fail
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The installation process involves downloading code from a third-party GitHub repository (BenedictKing/exa-search) which is not associated with a verified organization.
- [REMOTE_CODE_EXECUTION]: The command
npx skills add -g BenedictKing/exa-searchdownloads and executes external scripts on the host machine, posing a risk of arbitrary code execution from an untrusted source. - [PROMPT_INJECTION]: The skill ingests data from external web sources via the Exa API, creating a vulnerability to indirect prompt injection where instructions hidden in search results could influence agent behavior.
- Ingestion points: SKILL.md (via Exa API search results processed at runtime)
- Boundary markers: None identified in the skill instructions to delimit untrusted data.
- Capability inventory: Retrieval and processing of external semantic search results.
- Sanitization: No evidence of sanitization or filtering for retrieved content.
Recommendations
- AI detected serious security threats
Audit Metadata