
Docs Researcher
- Updated June 3, 2026
- amoscicki/aromatt
Manages a project knowledge base: initialize it once, then provide a topic to research and capture into project documentation. A developer uses it to keep a living, researched docs store alongside the codebase instead of scattering notes.
Key points
- Project knowledge base management
- Init to set up
- Topic-driven research
Docs Researcher by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add amoscicki/aromatt/plugin install docs-researcher@amoscicki-pluginsAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Last updated | June 3, 2026 |
|---|---|
| Repository | amoscicki/aromatt ↗ |
What it does
Set up and manage a project knowledge base, researching topics to populate and maintain project documentation.
README.md
docs-researcher
A Claude Code plugin that provides a documentation research agent. The agent searches official documentation, filters relevant information for the current task, and saves it to .claude/skills/project-knowledge-base/ as a reusable project-local knowledge base skill.
Purpose
This plugin enables Claude to build its own knowledge base for each project. Instead of repeatedly searching for the same documentation, Claude:
- Initializes the project knowledge base skill (if not exists)
- Checks
.claude/skills/project-knowledge-base/references/for existing documentation - If not found, uses the
docs-researcheragent to gather information - Saves filtered, task-relevant documentation for future sessions
- Updates the skill index for easy discovery
Installation
Copy the docs-researcher folder to your Claude Code plugins directory, or reference it in your dotfiles setup.
How It Works
Automatic Initialization
When docs-researcher runs in a new project, it automatically:
- Creates
.claude/skills/project-knowledge-base/structure - Initializes
SKILL.mdwith references index - Creates
references/directory for documentation - Optionally adds knowledge base section to project's
CLAUDE.md - Migrates any existing
.claude/knowledge/files (legacy format)
Knowledge Base Skill
The skill created in each project provides:
{project}/
└── .claude/
└── skills/
└── project-knowledge-base/
├── SKILL.md # Index with all references
└── references/
├── react-hooks.md
├── tanstack-router/ # Tree for large topics (>500 lines)
│ ├── _index.md
│ └── route-guards.md
└── ...
Project Setup
To enable the knowledge-building workflow in a project, add this to your project's CLAUDE.md:
## Knowledge Base
Ten projekt używa `.claude/skills/project-knowledge-base/`.
Przed kodowaniem zapoznaj się ze skillem.
The agent will add this section automatically if it doesn't exist.
Components
Agent: docs-researcher
Autonomous agent that:
- Initializes project knowledge base skill (Step 0)
- Validates research requests (technology, topic, context required)
- Checks existing references before researching
- Uses WebSearch/WebFetch for documentation
- Filters results for relevance
- Saves to standardized knowledge format
- Updates SKILL.md index
- Applies progressive disclosure for large documents (>500 lines)
Model: Haiku (fast, cost-effective for research tasks)
Skill: research-methodology
Provides the agent with:
- WebSearch query patterns for different technology domains
- Source prioritization (official docs first)
- Filtering criteria
- Document template and formatting rules
- Progressive disclosure guidelines
Script: init.js
Node.js script that creates the knowledge base skill structure:
- Creates
.claude/skills/project-knowledge-base/ - Initializes
SKILL.mdwith base template - Creates
references/directory
Knowledge File Format
Files are saved as .claude/skills/project-knowledge-base/references/{technology}-{topic}.md:
---
topic: "React useEffect Cleanup for Subscriptions"
technology: "react"
version: "18.x"
sources:
- https://react.dev/reference/react/useEffect
created: 2025-01-15
context: "Memory leak in subscription component"
---
Progressive Disclosure
When a reference exceeds 500 lines, it's automatically split into a tree:
references/tanstack-router.md (>500 lines)
↓ split to
references/tanstack-router/
├── _index.md # Overview + TOC
├── route-guards.md
├── data-loading.md
└── navigation.md
Migration
Projects with the legacy .claude/knowledge/ format are automatically migrated:
- Agent detects existing
.claude/knowledge/*.mdfiles - Copies each file to
.claude/skills/project-knowledge-base/references/ - Adds entries to SKILL.md index
- Does NOT delete originals - user verifies and removes manually
Directory Structure
docs-researcher/
├── .claude-plugin/
│ └── plugin.json
├── agents/
│ └── docs-researcher.md
├── scripts/
│ └── init.js
├── skills/
│ └── research-methodology/
│ ├── SKILL.md
│ └── references/
│ ├── query-patterns.md
│ └── document-template.md
└── README.md
License
MIT