Skip to content
Kitbay
Skill

ulw-research

NOASSERTIONUpdated on Nov 11, 202670,828GitHub stars

A shared skill from the oh-my-openagent project, used only when the user explicitly asks for deep research, that runs an exhaustive investigation with a team of read-only worker agents and ends in a cited, checked report.

What the skill teaches

A shared skill from the oh-my-openagent project, used only when the user explicitly asks for deep research, that runs an exhaustive investigation with a team of read-only worker agents and ends in a cited, checked report. The topic is split into research axes with at least one worker each, workers report leads and dead ends, and the coordinator sends new waves until leads run out. Every claim must rest on a tool observation such as a page or file read and is tracked in a claim graph and an observation log. The report, in PDF, Word, slides or HTML, goes through fidelity, meaning, layout, visual and proofreading checks. It was written for OpenCode, with a mapping for Codex.

Capabilities

  • Split a topic into research axes for workers
  • Track claims against tool observations
  • Write a cited report with charts and sources
  • Check the report for fidelity and layout