The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves arbitrary content from the web via the Exa API, creating a potential surface for indirect prompt injection.
Ingestion points: The scripts scripts/exa_search.py and scripts/exa_contents.py ingest titles, snippets, and full webpage text from external sources.
Boundary markers: While the instructions suggest selecting key highlights, there are no mandated strict delimiters (e.g., XML tags) for the untrusted content in the prompt.
Capability inventory: The skill has network capabilities limited to the Exa API and performs basic JSON parsing of the results.
Sanitization: The retrieved content is not explicitly sanitized for potential injection patterns before being presented to the agent.
[EXTERNAL_DOWNLOADS]: The documentation references official SDKs exa-py and exa-js from standard package registries for integration, which are well-known and appropriate for the skill's purpose.

exa-search