research-compendium

Archive primary sources, summarize each, and synthesize a self-contained research compendium.

Updated Aug 17, 2026
One-click install
npx skills add https://github.com/AMC-JTC/gbrain-1 --skill research-compendium-amc-jtc
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research-compendium
Source: https://github.com/AMC-JTC/gbrain-1/tree/main/plugin/skills/research-compendium
Command: npx skills add https://github.com/AMC-JTC/gbrain-1 --skill research-compendium-amc-jtc

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Deep research usually ends in scattered notes and lost sources. This Skill turns a research question into a permanent knowledge asset: every primary source archived verbatim, one summary per source, and a single synthesized compendium page that reads standalone. ## Core Features & Use Cases - Four-phase pipeline: find everything (brain-first, then web), archive each primary source under research/<topic>/sources/, write a strict 1:1 summary per source, then synthesize one compendium page with inline citations. - Depth dial: five idempotent levels from base synthesis up to exhaustive saturation, tracked in an index ledger so deepening a topic never redoes finished work. - Quality gates: cold-read accessibility check, mechanical fact-check of every claim against archived source spans, and a cross-modal eval before shipping. - Use Case: Ask for a definitive guide to spaced repetition and receive an archived corpus of 15-30 sources, per-source summaries with real numbers, and a self-contained compendium with a ranked reading manifest. ## Quick Start Research everything about spaced repetition, archive the sources, summarize each one, and write me a compendium.

Frequently Asked Questions about research-compendium

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a comprehensive research compendium from multiple sources?

State the topic and ask for a compendium. The skill decomposes the topic into angles, archives 15-30 quality primary sources verbatim, writes a 150-300 word summary per source, then synthesizes one self-contained compendium page with inline citations and a sources section.

What is the difference between research-compendium and data-research?

research-compendium produces prose knowledge synthesis backed by an archived source corpus. data-research extracts structured data into tracker pages with rows and fields. If the deliverable is a table or tracker, the request routes to data-research even when phrased as research.

Can I deepen an existing compendium without redoing the work?

Yes. Depth is a dial with five levels from base synthesis to exhaustive. The index page ledger tracks completed work, so bumping a level only adds the new layers such as full-text primaries, books, or counter-canon sources, never re-acquiring or re-summarizing existing pages.

Does the compendium archive sensitive or private sources verbatim?

Verbatim archiving is the default for public corpora like papers and articles, but it is gated by the user's retention posture. For personal, sensitive, or third-party-private material, the skill stores only the citation plus a summary and notes this in the index.

How are factual claims in the compendium verified?

At depth levels above base, every load-bearing claim is recorded in a claims ledger with the exact verbatim support span from an archived source. The fact-check skill then mechanically verifies each span exists in its cited source before the compendium ships.

When should I not use research-compendium?

Avoid it for quick single-question answers, structured data extraction into trackers, single-claim academic verification, or synthesizing content already in the brain. Those route to perplexity-research, data-research, academic-verify, or concept-synthesis respectively.