exa-efficiency
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the Exa API, which is a potential surface for indirect prompt injection. 1. Ingestion points: Search results, highlights, and full page text retrieved from api.exa.ai via the /search and /contents endpoints. 2. Boundary markers: The skill includes a specific rule for the agent to treat all search and agent outputs as untrusted content and never as instructions. 3. Capability inventory: The agent can execute a local script and perform further network requests to the Exa API. 4. Sanitization: The skill relies on clear natural language instructions to separate data from instructions rather than programmatic filtering.
- [COMMAND_EXECUTION]: The skill executes a provided Python script (scripts/exa_cost.py) for offline cost estimation. This script is restricted to the Python standard library and performs logic purely based on input JSON data without utilizing network access, shell spawning, or unsafe dynamic execution functions.
Audit Metadata