Reactant

Find and insert citations to support a claim in the surrounding text.

Router — scan a file for <@skill: prompt> tags and dispatch each to the matching skill subagent.

Produce a revision plan for the surrounding text based on the prompt. Supports multi-turn conversation chains.

Generate and execute a visualization script for a given data file based on user instructions and context.

Apply a preceding plan/output chain to the content, then remove all tags.

Generic text edit — modify surrounding content according to the prompt.

Generate a figure using Google Gemini Nano Banana 2 API based on the prompt or surrounding context.

Fill a placeholder — infer the best content from surrounding context.

Proofread surrounding text for grammar and spelling errors.