rusen-research logo

rusen-research

rusen-research

OtherClaude Codeby Rusen Birben

Summary

Research workflow skills for the rusen-brain Obsidian vault: paper triage and notes (with ML enrichment), citation traversal, literature surveys, idea exploration, vault primitives (vaultdb, obsidian-cli), and the /think command for Socratic learning sessions.

Install to Claude Code

/plugin install rusen-research@rusen-research

Run in Claude Code. Add the marketplace first with /plugin marketplace add rusen-archive/rusen-agent-skills if you haven't already.

README.md

rusen-research

Personal Claude Code plugin bundling a research workflow for the rusen-brain Obsidian vault.

The plugin is named rusen-research. The repo also acts as a single-plugin marketplace, so installation is a two-line claude plugin marketplace add + claude plugin install.

Skills

Research workflow (AI / ML reading)

| Skill | Description | |---|---| | lit-research | Orienting survey of a new research area — discover the landscape, cluster by method / finding / theory / problem framing, produce an executive overview. First-touch skill when entering a new topic. | | paper-triage | Fast SKIM/READ/DEEP/SKIP verdict for a paper from arxiv, Zotero, DOI, or a PDF path. | | paper-notes | Turn a paper (or a Zotero collection/tag) into Article Notes in the rusen-brain vault. For ML/AI papers, also produces an ML Context enrichment block (benchmarks vs SOTA, code, compute, follow-ups, criticisms). Pass mode=enrich-only to produce just the enrichment block without writing a note. | | lit-review | Synthesize a literature review from a set of papers into a vault note: thematic clusters, comparison, synthesis, gaps. | | citation-trace | Explore the citation graph around a seed paper — forward, backward, or both — with Zotero and vault cross-reference. | | idea-deep-dive | Explore the literature around a research question or hypothesis; cluster findings as supporting / contradicting / adjacent / tangential. |

Vault & tooling primitives

| Skill | Description | |---|---| | obsidian-cli | Interact with an open Obsidian vault via the obsidian CLI — read, create, search, manage notes, or develop plugins. | | vaultdb | Query, mutate, and traverse Obsidian vaults using vaultdb — a Rust CLI that treats markdown folders as a database with a citation graph. |

General

| Skill | Description | |---|---| | context-init | Explore an unfamiliar codebase in read-only mode and produce a concise onboarding report. |

Commands

Slash commands invoked explicitly with /<plugin>:<command>:

| Command | Description | |---|---| | /rusen-research:think | Enter Thinking Buddy mode — Socratic, no first-draft generation, explain-back required, predict-before-reveal. A friction-on learning mode designed to use AI to sharpen the user rather than to ship faster output. Optional argument: a topic to seed the session (e.g. /rusen-research:think B-trees vs B+ trees). Exit by saying ship mode or just answer. |

Shared design principles

Skills in the research workflow group follow strict conventions:

1. Always ask, never default. If the user hasn't specified a parameter, the skill asks before proceeding — it doesn't silently pick a default. This is an explicit top-of-skill invariant. 2. Canonical identifier is the Zotero citation key. Accept arxiv IDs, DOIs, and PDF paths; normalize internally. 3. Vault writes go through /vaultdb and /obsidian-cli. The research skills orchestrate; they don't reinvent vault machinery. 4. Body style matches the vault: encyclopedic prose with dense [[wiki-links]], not bullet dumps. Vault frontmatter respects the 3-tag schema (type/X + topic/Y + source/Z). 5. Cross-skill hand-offs are explicit. paper-triagepaper-notes; idea-deep-divecitation-tracelit-review. Skills name each other in their output "recommended next moves."

Layout

rusen-agent-skills/
├── .claude-plugin/
│   ├── plugin.json          # plugin manifest
│   └── marketplace.json     # one-plugin marketplace pointing at "./"
├── skills/
│   ├── citation-trace/SKILL.md
│   ├── context-init/SKILL.md
│   ├── idea-deep-dive/SKILL.md
│   ├── lit-research/SKILL.md
│   ├── lit-review/SKILL.md
│   ├── obsidian-cli/SKILL.md
│   ├── paper-notes/SKILL.md
│   ├── paper-triage/SKILL.md
│   └── vaultdb/SKILL.md
├── commands/
│   └── think.md             # /rusen-research:think — Thinking Buddy mode
├── README.md
└── LICENSE

Installing

# 1. Add this repo as a marketplace (one time)
claude plugin marketplace add https://github.com/rusenbb/rusen-agent-skills

# 2. Install the plugin
claude plugin install rusen-research@rusen-research

# 3. Restart Claude Code

Skills will appear under the rusen-research: namespace (e.g. rusen-research:paper-triage).

To install from a local clone instead:

git clone https://github.com/rusenbb/rusen-agent-skills ~/code/rusen-agent-skills
claude plugin marketplace add ~/code/rusen-agent-skills
claude plugin install rusen-research@rusen-research

Changes to skill files take effect in the current Claude Code session without restart.

Adding a skill

1. Create a new directory under skills/ named after the skill (kebab-case). 2. Add a SKILL.md with at least name and description in the frontmatter. 3. Keep the description specific — it's what Claude reads to decide when to invoke the skill.

Adding a command

1. Create a markdown file under commands/ named after the command (kebab-case). 2. Add YAML frontmatter with at least description. Optional: argument-hint for a placeholder shown in the picker. 3. Body is the prompt template. Use $ARGUMENTS to interpolate user-supplied arguments. 4. Invoke as /rusen-research:<name> after claude plugin marketplace update rusen-research + restart.

Changelog

  • 0.2.0 — Add /rusen-research:think command (Thinking Buddy mode — Socratic, no first-draft generation, explain-back required).
  • 0.1.0 — Repackaged as a Claude Code plugin (was: 11 loose skills installed via symlink). Skills now live under skills/. Two retired: arxiv-presentation (unused), ml-paper-context (folded into paper-notes as the ML Context block + enrich-only mode).

Related plugins

Browse all →