Skip to content
SkillModel-led1,242 tokens in use

autoresearch

A budgeted research loop that gathers cited sources, drafts a dossier, and files it into an Obsidian vault only after you review it.

16,653 stars on the repository MIT LicenseAgriciDaniel/claude-obsidian · skills/autoresearch
claude plugin marketplace add AgriciDaniel/claude-obsidian claude plugin install claude-obsidian@agricidaniel-claude-obsidian
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing autoresearch costs

83tokens idle
1,242tokens in use
1,933tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Excellent
  • Frontmatter12/15
  • Structure19/25
  • Examples12/15
  • Output15/15
  • Size10/15
  • Resources15/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 22

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity84
  • Maintenance100
  • Community56
  • Maturity28
  • Momentum40

autoresearch runs a bounded research loop for a topic you name. It confirms the question, the approved domains and the budget with you, then searches in rounds, records each source with its authority and date, and extracts claims with evidence locators.

The result is a draft dossier. Nothing becomes canonical vault knowledge until you review it and approve a transaction, and merging into existing concept or entity pages is a second, separate approval.

What You Get

  • A cited research dossier drafted from official and primary sources first
  • Source and claim ledger records with authority, freshness and independence keys
  • A per-round budget report and a stated stop reason, including coverage gaps
  • A reviewed transaction bundle you can inspect before anything touches the vault
  • A separate proposal for merging findings into existing concept and entity pages

Common Use Cases

  • Researching a technology or product and filing a sourced summary into a personal wiki
  • Checking a claim against a counter-position before writing about it
  • Building a first-pass dossier on a topic with a hard cap on searches and pages

What Makes It Unique

Most research agents write straight into notes. autoresearch keeps research and merging apart: findings stay a draft dossier, high-risk claims need two independent sources, contradictions are preserved, and page creation is tied to an index update in one inspected, recoverable transaction. It also treats web pages, snippets and worker drafts as untrusted evidence and ignores instructions embedded in them.

Under The Hood

The skill is instructions plus a default research program in a references file. It tells the agent to search and fetch public web pages, but only after you approve network access and domains; without consent it reads just the vault and sources you supply. It runs the plugin’s Python core script through the shell to inspect and apply a transaction bundle, and optionally to create a Git checkpoint when you ask. Vault files are written only by that core, after you approve the plan hash. Parallel helper agents may search and draft but never write to the vault.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers