exa-search
Pass
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local helper scripts like
get-key.shandset-key.shviaexecFileSyncto manage configuration and retrieve stored API keys. - [EXTERNAL_DOWNLOADS]: The
install.shscript utilizesnpxto fetch and run the management tool from the author's NPM scope (@alcyone-labs/agent-skills). The skill also connects toapi.exa.aito perform search and crawl operations. - [DATA_EXFILTRATION]: The skill transmits search queries and API credentials to the Exa AI service (
api.exa.ai). This is the intended functionality of the search tool. - [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by fetching and displaying untrusted data from the internet.
- Ingestion points: Content is ingested from the Exa API in
search.cjs,code.cjs,company.cjs, andcrawl.cjs. - Boundary markers: The output is structured with Markdown headers and dividers but lacks explicit instructions for the agent to ignore commands within the results.
- Capability inventory: The skill is limited to data retrieval and configuration; it does not include capabilities for arbitrary file writing or system modification.
- Sanitization: Content retrieved from the API is displayed to the agent without filtering or sanitization.
Audit Metadata