wiki-research
Skill21 starsChanged 4 months agoMITView on GitHub
Autonomously research a topic via multi-round web search, synthesize findings, and file structured results into the Obsidian wiki. Use this skill when the user says "/wiki-research [topic]", "research X", "find everything about Y", "do a deep dive on Z", "autonomous research on X", or wants comprehensive, web-sourced knowledge on a topic filed directly into their wiki.
- Reads credentials
We split this skill into the rules it gives an agent
16 rules. About 0 are common, 11 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Read $VAULT/_system/index.md to understand what's already in the wiki — don't re-research things the wiki covers well
- Very specific: line 19
Read $VAULT/_system/hot.md — surfaces recent context and active threads
- Very specific: line 20
Check $VAULT/_system/tags.md — use only the tags in the whitelist when writing new notes
- Very specific: line 29
If references/research-config.md exists in the vault, read it and apply any rules it defines:
- Distinctive: line 35
If the file doesn't exist, proceed with defaults.
- Distinctive: line 61
Run up to 5 targeted searches specifically addressing these gaps.
- Distinctive: line 63
Add findings to your working set.
- Distinctive: line 69
If major contradictions remain unresolved:
- Distinctive: line 70
Run one final targeted pass (2-3 searches) to find authoritative resolution
- Distinctive: line 71
If resolution is impossible, flag the contradiction explicitly in the synthesis page
- Distinctive: line 73
If contradictions are minor or the topic feels well-covered after Round 2, skip additional searching and proceed to filing.
- Distinctive: line 75
Halt condition: Stop when depth is achieved or 3 rounds are complete — do not loop indefinitely.
- Distinctive: line 163
After filing all pages:
- Distinctive: line 164
Every concept page should link to at least 2 source pages
- Distinctive: line 165
Every source page should link to the concept pages it informed
- Very specific: line 168
Check index.md for existing pages on the same topics — merge into existing pages rather than creating duplicates.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Wiki Research — Autonomous Multi-Round Research
You are running an autonomous research loop on a topic, synthesizing what you find, and filing the results into the Obsidian wiki as permanent knowledge.
Before You Start
Resolve the vault path (precedence, highest first): the OBSIDIAN_VAULT_PATH environment variable if set, else a .env in the current working directory (vault-scoped), else ~/.obsidian-wiki/config (global default). Set VAULT to that OBSIDIAN_VAULT_PATH.
- Read
$VAULT/_system/index.mdto understand what's already in the wiki — don't re-research things the wiki covers well - Read
$VAULT/_system/hot.md— surfaces recent context and active threads - Check
$VAULT/_system/tags.md— use only the tags in the whitelist when writing new notes
Source Verification (mandatory before filing any note):
For every claim you intend to write into a wiki page, you must be able to point to a URL you successfully fetched with WebFetch. Search snippets alone are not sufficient. Drop any claim whose source page cannot be verified.
Confirm the research topic with the user if it's ambiguous. Then proceed.
Research Configuration (optional)
If references/research-config.md exists in the vault, read it and apply any rules it defines:
- Source preferences (e.g., prefer academic sources, avoid certain domains)
- Domains to skip
- Confidence scoring adjustments
- Topic-specific constraints
If the file doesn't exist, proceed with defaults.
Round 1 — Broad Survey
Goal: Get a wide map of the topic.
- Decompose the topic into 3-5 distinct angles (e.g., for "vector databases": what they are, when to use them, leading implementations, trade-offs, production gotchas)
- For each angle, run 2-3
WebSearchqueries using varied phrasing - For the top 2-3 results per angle, use
WebFetch(ordefuddle <url>if available — cleaner extraction) to get content - From each fetched page, extract:
- Key claims — what the source explicitly states
- Concepts — ideas, terms, frameworks introduced
- Entities — tools, people, organizations mentioned
- Contradictions — places where sources disagree with each other
Track what's covered and what's missing as you go.
Round 2 — Gap Fill
Goal: Close the holes left by Round 1.
Review what Round 1 produced:
- What questions did sources raise but not answer?
- Where do sources contradict each other?
- Which angles got thin coverage?
Run up to 5 targeted searches specifically addressing these gaps. Prefer primary sources, official documentation, and authoritative analyses over link aggregators.
Add findings to your working set. Update the contradiction list.
Round 3 — Synthesis Check
Goal: Resolve contradictions; confirm depth is sufficient.
If major contradictions remain unresolved:
- Run one final targeted pass (2-3 searches) to find authoritative resolution
- If resolution is impossible, flag the contradiction explicitly in the synthesis page
If contradictions are minor or the topic feels well-covered after Round 2, skip additional searching and proceed to filing.
Halt condition: Stop when depth is achieved or 3 rounds are complete — do not loop indefinitely.
Filing — Write Wiki Pages
Organize all findings into wiki pages across four output areas:
1. sources/ — One page per major reference
For each significant source (typically 4-8 pages total):
---
title: >-
<Source title>
category: references
tags: [<2-4 domain tags>]
sources:
- "<URL>"
source_url: "<URL>"
created: <ISO-8601 timestamp>
updated: <ISO-8601 timestamp>
summary: >-
<1-2 sentences describing what this source covers, ≤200 chars>
provenance:
extracted: 0.X
inferred: 0.X
ambiguous: 0.X
---Body: title, URL, what it covers, key claims (with provenance markers), limitations.
2. concepts/ — One page per substantive concept
For each significant concept surfaced across sources:
Standard concept frontmatter + body. Link concepts to each other and to source pages.
3. entities/ — Tools, organizations, people
For each significant entity encountered (tools, libraries, companies, key authors):
Standard entity frontmatter. Link back to concepts that use the entity and sources where it appears.
4. synthesis/Research: [Topic].md — Master synthesis
The primary output: a structured synthesis of everything found.
---
title: >-
Research: <Topic>
category: synthesis
tags: [<3-5 domain tags>, research]
sources: [<list of source URLs or page paths>]
created: <ISO-8601 timestamp>
updated: <ISO-8601 timestamp>
summary: >-
Synthesis of <N>-round research on <topic>. Covers <core findings in ≤200 chars>.
provenance:
extracted: 0.X
inferred: 0.X
ambiguous: 0.X
---
# Research: <Topic>
## Overview
<2-4 sentence executive summary of what the research found>
## Key Findings
<Bulleted list of the most important claims, each with a [[source page]] citation>
## Core Concepts
<Links to concept pages created, with one-line descriptions>
## Entities & Tools
<Links to entity pages, with one-line descriptions>
## Contradictions & Open Questions
<Where sources disagree or where the research hit limits>
## Sources Consulted
<Linked list of all source pages>Cross-linking
After filing all pages:
- Every concept page should link to at least 2 source pages
- Every source page should link to the concept pages it informed
- The synthesis page should link to all concept, entity, and source pages produced
Check index.md for existing pages on the same topics — merge into existing pages rather than creating duplicates.
Update Tracking Files
.manifest.json — Add a research entry:
{
"type": "research",
"topic": "<topic>",
"researched_at": "TIMESTAMP",
"rounds_completed": 3,
"sources_fetched": N,
"pages_created": ["..."],
"pages_updated": ["..."]
}index.md — Add all new pages under their respective sections.
log.md — Append:
- [TIMESTAMP] WIKI_RESEARCH topic="<topic>" rounds=N sources_fetched=N pages_created=Mhot.md — Update Recent Activity with the research topic and core finding. Update Active Threads if this is ongoing. Update updated timestamp.
Quality Checklist
- 3 rounds completed (or halted at sufficient depth)
- Synthesis page exists at
synthesis/Research: [Topic].md - Source pages written for major references
- Concept and entity pages written for significant items
- Contradictions flagged in synthesis page
- All pages cross-linked
index.md,log.md,hot.md,.manifest.jsonupdated
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- chobizzy/llm-wiki★ 280% same
Wrote their own version · 1
- SkinnnyJay/wiki-llm★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- claude-history-ingestSkill
- codex-history-ingestSkill
- cross-linkerSkill
- data-ingestSkill
- graph-colorizeSkill
- ingest-urlSkill
- llm-wikiSkill
- tag-taxonomySkill
- wiki-captureSkill
- wiki-dashboardSkill
- wiki-exportSkill
- wiki-ingestSkill
- wiki-lintSkill
- wiki-querySkill
- wiki-rebuildSkill
- wiki-setupSkill
- wiki-statusSkill
- wiki-synthesizeSkill
- wiki-updateSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

