The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches arbitrary data from the web, creating a potential surface for indirect prompt injection attacks where malicious instructions hidden on websites could influence the agent's behavior.
Ingestion points: The skill retrieves web search results, full-page text, and synthesized answers in scripts/exa_search.py, scripts/exa_contents.py, scripts/exa_research.py, and scripts/exa_agent.py.
Boundary markers: The skill does not implement specific delimiters or warnings to isolate untrusted web content from the agent's primary instructions.
Capability inventory: The skill has the capability to perform network requests to the Exa API and return extensive text data to the agent context. It does not have local file system write access or arbitrary command execution capabilities.
Sanitization: No sanitization or filtering is performed on the retrieved web content to detect or neutralize embedded instructions.
[DYNAMIC_EXECUTION]: The testing utility scripts/test_offline.py uses importlib.import_module to dynamically load the skill's functional scripts for validation purposes, which is a standard practice in test environments.
[COMMAND_EXECUTION]: The script scripts/test_exa.py uses subprocess.run to execute the skill's own Python scripts to verify CLI functionality during testing.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the standard requests Python library to communicate with the Exa API.

exa-search