The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes untrusted data from external websites, creating a surface for indirect prompt injection. Findings from the open web are synthesized into reports, which could allow malicious content from third-party sites to influence the agent's output.
Ingestion points: Web content is fetched via firecrawl_search, firecrawl_scrape, web_search_exa, and crawling_exa tools as described in the workflow in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or explicit 'ignore' instructions when the agent processes the retrieved content.
Capability inventory: The skill possesses network read capabilities (searching/scraping) and file write capabilities for saving research reports.
Sanitization: There are no explicit sanitization steps or filtering mechanisms defined for the scraped content before it is processed by the model.
[EXTERNAL_DOWNLOADS]: The skill interacts with external web services to retrieve research data. It utilizes Firecrawl and Exa, which are established tools for web crawling and search functionality.

deep-research