Maintain a project's Research layer in an Obsidian vault — scaffold a research note from a question, spawn experiment tasks with backlinks, roll experiment findings up into the research note, sync open questions, and on conclusion graduate durable findings into Topics. Use when the user wants to start a piece of research, add an experiment to a research question, update a research note from its experiments, review open questions, or conclude a research question. Sibling of obsidian-topic-main...
Scanned 9/5/2026
Install to Claude Code
npx -y skills add channingwalton/dotfiles --skill obsidian-research-maintainer --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Obsidian Research Maintainer?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/channingwalton-obsidian-research-maintainer-dotfiles)More formats (shields.io, HTML) on the badges page.
---
name: obsidian-research-maintainer
description: Maintain a project's Research layer in an Obsidian vault — scaffold a research note from a question, spawn experiment tasks with backlinks, roll experiment findings up into the research note, sync open questions, and on conclusion graduate durable findings into Topics. Use when the user wants to start a piece of research, add an experiment to a research question, update a research note from its experiments, review open questions, or conclude a research question. Sibling of obsidian-topic-maintainer; assumes projects under the vault have Research/, Tasks/, Topics/ (and usually Events/) sub-folders.
---
# Obsidian Research Maintainer
Maintains the **Research** layer of an Obsidian project. A research note holds one research question end to end: problem, hypotheses, links to experiments, a rolling summary of findings, and open questions. Experiments are ordinary **Tasks** (`task-type: experiment`) that link back to their research note — no separate experiment structure exists or should be invented. Supporting documents are captured in **Events** and linked from the research note.
For requested changes whose scope and content are established, write directly and show the result. Ask only when a substantive choice remains unresolved, such as competing interpretations of evidence or a new research direction. Existing authorisation carries through the workflow; an audit-only request produces findings.
## Vault model & conventions
- Vault root is typically `~/Documents/Notes`; projects live under `Projects/<Name>/` with `Research/`, `Tasks/`, `Topics/`, and usually `Events/`.
- `Research/Research.md` is the index hub for a project's research notes; the project note links it in its footer line.
- **Every note's H1 is a self wikilink** (`# [[Note Title]]`) followed by a parent link line — `[[Research]]` for research notes.
- British spelling. Generate any timestamp with the shell `date` command, never a guess.
- **Research note** shape (template: `Vault Metadata/Templates/Inputs/research.md`) — no date prefix in the filename; named by a short form of the question:
```
---
status: open # open | concluded
projects:
- "[[<Project>]]"
dateCreated: <ISO8601>
dateModified: <ISO8601>
tags:
- research
---
# [[<Title>]]
[[Research]]
## Problem / ## Hypotheses / ## Experiments / ## Findings / ## Open Questions / ## Documents
```
- **Experiment task** = a normal Task note (dated filename, task frontmatter) plus:
```
task-type: experiment
research:
- "[[<Research Note>]]"
```
and a `[[<Research Note>]]` backlink line under its H1. The task keeps its own Current State, method, Results, Decision Log; links are one-way in prose (task → research), while the research note's **Experiments** section lists its experiment tasks with a one-line description each.
## Division of content — the important rule
- **Experiment task**: everything specific to that run — setup, protocol, raw results, per-experiment decisions, next-session notes.
- **Research note**: only what spans experiments — the question, hypotheses, distilled findings, research-level open questions.
- When in doubt, leave detail in the task and link it. A research note that duplicates a results table has gone wrong.
## Workflows
### Start a research question
1. Ask for the question if not given; use an agreed short title or choose one that preserves the question's meaning.
2. Create `Research/<Title>.md` from the template shape above (shell `date` for timestamps). Write the Problem section from the user's phrasing. Record agreed hypotheses; label newly drafted hypotheses as provisional and ask only if choosing between them changes the research scope or method.
3. Add the note to `Research/Research.md`'s list. Create `Research/` and `Research.md` first if the project lacks them, and add the footer link to the project note.
### Add an experiment
1. Create a Task note (dated filename, matching the project's task conventions) with `task-type: experiment` and the `research:` frontmatter + H1 backlink.
2. List it under the research note's **Experiments** section with a one-line description. Bump `dateModified`.
### Roll up findings
1. Read each linked experiment task's Current State / Results / Decision Log.
2. Update **Findings** from established results: short bolded claims, each attributable to an experiment, with an `*Updated: [[<date>]]*` line. Put cross-experiment questions in the research note and record answers in Findings before resolving research questions. If the synthesis requires an unresolved interpretation, ask about that claim before recording it as a finding.
3. Bump `dateModified` and show the changes. The experiment tasks remain unchanged during roll-up except for a missing backlink; use `task-note-update` if the user also requests experiment-task updates.
### Conclude a research question
1. When the user asks to conclude the research, record whether it is answered or abandoned from the established context; ask only if that disposition is unclear. Set `status: concluded` and add a dated **Conclusion** section summarising the outcome.
2. **Graduate durable findings into Topics**: use `obsidian-topic-maintainer` for Topics updates already within the requested scope. Otherwise present candidate hubs and destinations for approval. Write in topic-hub voice — concept prose, no experiment enumeration, `## See also` links.
3. The research note stays in `Research/` as the archived record; the Topic carries the knowledge forward.
### Audit
On request, check across a project: every `task-type: experiment` task has a resolving `research:` link and H1 backlink; every research note's Experiments list matches the tasks that claim it; every open research note has been updated since its newest experiment activity (flag stale ones); `Research.md` lists every research note. Report findings; apply fixes when requested, using any authorisation already given.
## Notes
- Do not create per-research sub-folders, an Experiments/ folder, or any structure beyond `Research/` — experiments are Tasks, documents are Events.
- Generalises across projects: point it at any `Projects/<Name>/` and offer to scaffold `Research/` if absent.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!