Deep-Research-skills

Why we included this project

Deep research usually means a long session of back-and-forth prompting, and the results are only as good as the last instruction you gave. This skill packages that into a two-phase workflow: it first builds an outline of the items you want covered and the fields to collect for each, then sends parallel web-search agents to investigate and merges everything into one markdown report. The part that earns its keep is the human check between phases, you review and extend the outline before the deep investigation starts, so the scope stays yours instead of drifting. That fits literature surveys, tool and framework comparisons, competitor analysis, and due-diligence work where you want consistent, structured output across many sources. It installs as skills and agents for Claude Code, OpenCode, and Codex, with English and Chinese versions, so a team can standardize on one research process and share it.