/neuroflow:ideation¶
The very beginning of a research project.
/ideation helps you brainstorm a research question, explore the literature, formalize an existing idea into a project definition, or produce a project proposal document. It's the natural next step after /neuroflow for a new project.
When to use it¶
- You have a vague idea and need to think through research questions
- You want to search what is already known before committing to a question
- You have an idea and want to sharpen it into a concrete, testable research question
- You want to produce a written project proposal
- You have a set of papers and want a structured, critic-checked literature review
Entry points¶
When you run /neuroflow:ideation, Claude asks which mode applies:
You have a vague idea. Claude helps you narrow it down to a testable research question.
What Claude does: - Asks you to describe your idea in any form - Helps identify key variables (independent, dependent, confounds) - Identifies population and modality - Formulates a clear hypothesis statement
Output: research-question.md saved to .neuroflow/ideation/
Example:
You: I'm interested in how background noise affects EEG-based attention measures.
Claude: Let me help you narrow this down. What kind of attention are you
measuring — sustained, selective, or divided?
You: Selective attention — I want to look at P300 amplitude differences.
Claude: Good. Here's a draft research question:
"Does white noise background (65 dB) reduce P300 amplitude during
a visual oddball task compared to silence in healthy adults?"
Key variables:
- Independent: noise condition (noise vs. silence)
- Dependent: P300 amplitude and latency
- Confounds: hearing threshold, fatigue, time of day
Search PubMed and bioRxiv simultaneously using the inline search protocol (no sub-agent spawned), with CrossRef, Semantic Scholar and arXiv as fallbacks.
No credentials required
PubMed and bioRxiv search works out of the box — no setup needed.
What Claude does: - Runs your topic on both PubMed and bioRxiv, then CrossRef / Semantic Scholar / arXiv when bioRxiv coverage is thin - Tries synonym and broader/narrower queries if results are thin - Returns a deduplicated list with ⚠️ preprint and 🔒 paywall markers - Saves a metadata stub per paper; each DOI is labelled with the check actually done (for example "from the PubMed record" or "resolves on 2026-10-07"), never "verified" - Downloads only the papers you pick, and only open-access copies (preprint servers, open repository copies, publisher open access) — never Sci-Hub. For the rest it tells you which PDFs to get through your library and where to save them - Offers follow-up: literature review, save as markdown, synthesize, or watch the query
Keep following a topic
Choose watch to pin a query in .neuroflow/ideation/watch.md. Claude recommends a free PubMed alert ("Create alert") and bioRxiv / medRxiv subject alerts; when a pinned query has not been checked for a week, /ideation offers to re-run it and shows only the new papers — they join your paper set only if you pick them.
Zotero (optional)
With a Zotero MCP server connected, Claude searches your library first and skips papers you already have. It writes to Zotero (collections, notes, tags) only when you confirm each write.
Output: literature-[topic]-[date].md saved to .neuroflow/ideation/
Example output:
PubMed results (5 papers)
─────────────────────────
**N2 and P300 in auditory attention** (2023) — Smith et al.
*NeuroImage* | DOI: 10.1016/j.neuroimage.2023.001
Shows P300 amplitude reduces with cognitive load in selective attention tasks.
bioRxiv results (2 papers)
──────────────────────────
**Noise effects on ERP** (2024) — Jones et al.
*bioRxiv* | DOI: 10.1101/2024.001
⚠️ PREPRINT
Summary: The literature consistently shows P300 attenuation under high
cognitive load. White noise as a stressor is understudied — gap identified.
You have an idea. Claude sharpens it into a precise, testable research question.
Similar to Brainstorm but focused: Claude takes your existing idea and pressure-tests it for clarity, testability, and novelty.
Produce a structured project proposal document.
Sections: - Research question - Background (from your literature) - Hypothesis - Planned methods - Population and modality - Rough timeline
Output: proposal-[date].md saved to .neuroflow/ideation/
Run 12 analytical protocols — intake, contradictions, gaps, timeline, methods, citation network, review prose, devil's advocate, theory, variables, plain language, research agenda — on the papers in .neuroflow/ideation/papers/ (PDFs or metadata stubs).
What Claude does:
- Confirms the paper list with you, then runs the literature-review agent
- The agent checks every protocol with a critic pass (its rubric, countable items first) — a self-check, labelled as such — and saves each protocol as a checkpoint, so an interrupted review resumes where it stopped
- Shows a status table (approved or halted per protocol); on request, a separate critic agent re-checks a protocol against the paper files
Output: literature-review-[date].md in .neuroflow/ideation/, with per-protocol checkpoints in literature-review-[date]/
Integration reminders¶
PubMed / bioRxiv — no credentials required, works out of the box.
Miro — if you mention Miro or ask to visualize a mind map, Claude checks whether Miro tools are available in the session. If not, it shows how to add Miro yourself with claude mcp add in a separate terminal (the /setup Step 2 instructions); it never asks for your token.
Files read and written¶
| Direction | Files |
|---|---|
| Reads | .neuroflow/project_config.md, .neuroflow/flow.md, .neuroflow/ideation/flow.md, .neuroflow/ideation/watch.md, ~/.neuroflow/user.yaml |
| Writes | .neuroflow/ideation/ (including papers/, watch.md, literature-review-[date].md), .neuroflow/ideation/flow.md, .neuroflow/sessions/YYYY-MM-DD.md, ~/.neuroflow/user.yaml (your Zotero answer) |
Related commands¶
/grant-proposal— write a formal grant application after ideation/experiment— design the paradigm once the research question is clear/data-analyze— run analysis with results tied back to your hypothesis