Madosho research

Skill hogu-dev/madosho/skills/madosho-research

a headless retrieval backend. Serves traditional RAG via a tool server or OpenAI shim, drive it from agents over MCP, CLI, or skills, run autonomous research subagents, and rate and compare each pipeline build step in an optional frontend

Install

npx -y skills add hogu-dev/madosho --skill madosho-research

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when you want an autonomous research pass over a madosho RAG corpus that returns a finished, cited report. Composes madosho-search and dispatches a subagent that follows the bundled autonomous.md playbook to gather evidence and write the report.

SKILL.md

2.2 KB, as published. Nobody here has run it

Autonomous research over a madosho corpus

Use this when the goal is a finished, cited report -- not interactive tool-driving. It builds on madosho-search (the same madosho-cli tools) and runs the research playbook as a self-contained pass that hands back a report.

How it runs

Make sure the tools are reachable (see madosho-search for env vars). You need to know the corpus name; madosho-cli list-corpora --json if unsure.
Dispatch a subagent whose instructions are the contents of the bundled autonomous.md (in this skill's directory). Read that file and pass its text as the subagent's instructions -- that file is the research judgment, written down. (This "dispatch a subagent with these instructions" step maps to each host's own subagent mechanism; it does not depend on any host-specific agent-config file.)
Give the subagent the research question and the corpus name, and tell it to use the madosho-search tools (madosho-cli search / search-doc / get-doc / list-corpora / list-documents / list-pipelines, always with --json).
The subagent runs the bounded loop in autonomous.md -- orient, focused search rounds, get-doc when it needs a whole document, stop when it can answer -- and returns a markdown report with inline citations.
Hand the report back to the user. Do not re-summarize away its citations.

Why a subagent

The playbook is the human's judgment written down, so it attaches only to the machine doing the autonomous work. The parent agent stays free to keep talking to the user while the subagent gathers and writes.

Server-side alternative

madosho can also run this exact playbook on its own server. To trigger that instead of dispatching a local subagent, see research_trigger.py in this pack (it POSTs to madosho's /corpora/{id}/research and polls for the report). That path needs an LLM provider configured on madosho's side.