Hey all,
I’m working on creating a swarm of deep agents and want to leverage skills to avoid bloating the context window.
Specifically, I have agents that perform research tasks, and I’d like them to be able to generate PDFs programmatically using a skill file (similar to how Anthropic’s SKILL.md files work to provide specialized instructions and best practices).
Is there a way to create a custom PDF generation skill that my agents can reference? If not, what alternative approaches would you recommend for enabling agents to create PDFs without bloating the context?