Agents¶

Agents are autonomous subprocesses launched by commands when deeper, focused work is needed.

Unlike commands (which interact conversationally with you), agents operate semi-autonomously on a specific task — searching a database, auditing a folder, running a check. A command invokes an agent, the agent does the work, and the result flows back to you.

Available agents¶

`scholar`¶

Academic literature research specialist.

Searches PubMed then bioRxiv sequentially (with CrossRef / Semantic Scholar fallbacks) for a given topic and returns a clean, structured list of results.

Invoked by: /neuroflow:ideation (Explore literature mode)

What it does:

Runs the query sequentially: PubMed first, then bioRxiv, then fallbacks one at a time if needed
If results are thin, generates 2–3 alternative queries (synonyms, narrower/broader terms)
Deduplicates results across sources
Downloads open-access full text in batches of 2; skips paywalled papers (saves metadata stub)
Returns results in a structured format with markers:
⚠️ PREPRINT — bioRxiv papers that have not been peer-reviewed
🔒 PAYWALLED — papers without open-access full text

Output format:

PubMed results
──────────────
**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
───────────────
**Noise effects on ERP** (2024) — Jones et al.
*bioRxiv* | DOI: 10.1101/2024.001
⚠️ PREPRINT — White noise as stressor reduces P300 in healthy adults.

Summary: P300 attenuation under high cognitive load is consistent in the
literature. White noise as a specific stressor is understudied — a gap exists.

Follow-up actions after results:

Action	What happens
`"download"`	Fetch full text for open-access papers (skips paywalled)
`"save"` / `"md"`	Save as `literature-[topic]-[date].md` in `.neuroflow/ideation/`
`"summarize"`	Deeper synthesis: main findings, methodological patterns, contradictions

Rules: - Never fabricates papers, authors, or DOIs - If a DOI cannot be verified, it is marked as unverified - PubMed and bioRxiv results are always presented separately

No credentials required

PubMed and bioRxiv search is handled by the paper-search-mcp-nodejs server — no credentials needed.

`sentinel`¶

Project coherence guard.

Audits .neuroflow/ for internal consistency and drift. Called by the /neuroflow:sentinel command when in a project repository (not a plugin repository).

Invoked by: /neuroflow:sentinel (when .neuroflow/ exists)

What it checks:

flow.md completeness — files listed vs files on disk
Timestamp drift — stale flow.md vs recent file activity
Broken references in reasoning/ JSON files
Phase consistency — active phase vs session logs vs folder activity
Preregistration drift — planned analyses vs what was actually done
Plugin version sync — project_config.md vs current plugin.json
Subfolder name validation — no unrecognized or skill-named folders
CLAUDE.md neuroflow reference check

See /sentinel for full documentation.

`sentinel-dev`¶

Plugin development coherence guard.

Audits the neuroflow plugin repository itself for structural consistency. Called by the /neuroflow:sentinel command when run inside the plugin repo (where .claude-plugin/plugin.json exists).

Invoked by: /neuroflow:sentinel (when .claude-plugin/plugin.json exists)

What it checks:

Command folder names vs frontmatter name: fields
Skill folder names vs SKILL.md frontmatter
README tables vs actual files
Version sync between plugin.json and all references
Dead references (links to files that don't exist)
Command frontmatter completeness

Phase agents¶

Specialist autonomous subprocesses, one per research phase.

Each phase agent has deep domain knowledge scoped to its phase. It operates with a plan-first, confirm-before-executing discipline — it drafts a plan, shows it to you, and only proceeds after confirmation.

Agent	Phase	What it does
`ideation`	ideation	Crystallises research questions via brainstorm, literature explore, formalise, or proposal modes
`grant-proposal`	grant-proposal	Structures proposals section by section for a target funder (NIH, ERC, Wellcome, MRC)
`experiment`	experiment	Paradigm design (PsychoPy), recording setup, and instrument configuration for EEG, fMRI, eye-tracking, ECG
`tool-build`	tool-build	Spec-first design and implementation of acquisition, real-time, LSL, BCI, and analysis pipeline tools
`tool-validate`	tool-validate	Timing, marker integrity, output format, and edge-case testing; writes validation plan before running any tests
`data`	data	Inventory → BIDS validation → conversion sequence; confirms modality before touching anything
`data-preprocess`	data-preprocess	Modality-aware preprocessing pipeline (EEG, fMRI, ECG, eye-tracking); documents all parameters before running
`data-analyze`	data-analyze	Statistical analysis — ERPs, time-frequency, connectivity, decoding, GLM; audits assumptions and applies multiple-comparison correction
`review`	review	Reads a colleague's paper and produces a structured referee report calibrated to the target journal; delegates to the `review-neuro` skill
`notes`	notes	Captures freeform input without interruption; reformats into a structured document only when asked
`write-report`	write-report	Synthesises `.neuroflow/` memory into a structured report for any phase or the full project
`brain-build`	brain-build	Spec-first design of neuron models and network topology for NEURON, Brian2, NetPyNE, NEST, tvb-library
`brain-optimize`	brain-optimize	Plans parameter sweeps or data-fitting runs; selects the right algorithm (grid, differential evolution, Bayesian, BluePyOpt)
`brain-run`	brain-run	Configures and executes simulation runs, sanity-checks outputs for silence, runaway activity, or NaN values; supports HPC job submission

How agents differ from commands¶

	Commands	Agents
Invoked by	You (directly)	Commands (programmatically)
Interaction style	Conversational	Autonomous task execution
Scope	Phase-level work	Focused sub-task
Output	Phase subfolder in `.neuroflow/`	Result returned to calling command