Skills¶

Skills are structured domain knowledge that Claude loads automatically when working in a specific phase.

Skills are not commands you run — they are context files that Claude reads in the background. When you run /neuroflow:data-preprocess, Claude automatically loads the neuroflow:phase-data-preprocess skill, which gives it neuroscience-specific guidance for that phase: what pipeline steps to apply, what domain conventions to follow, what common pitfalls to avoid.

How skills work¶

When a command runs, it starts by reading its associated skill. For example:

/neuroflow:data-preprocess
  → reads: neuroflow:phase-data-preprocess
  → reads: neuroflow:neuroflow-core (lifecycle rules)
  → reads: project_config.md and flow.md
  → starts the preprocessing pipeline

Skills give Claude phase-specific expertise without you having to instruct it manually.

Available skills¶

Core skills and protocols¶

Skill	What it does
`neuroflow:neuroflow-core`	Core rules and lifecycle for all commands and agents — `.neuroflow/` folder spec, the `project_config.md` contract, shared formats (decision logs, integrity markers, merge safety, sharing tiers), command lifecycle and frontmatter standard. Ships `scripts/scaffold.py` (used by `/neuroflow`) and `scripts/migrate.py` (used by `/migrate`)
`neuroflow:worker-critic`	Worker-critic loop protocol — a worker agent and a critic agent across up to 3 revision cycles; used by `/paper` and `/poster`
`neuroflow:autoresearch-protocol`	The improvement-loop protocol — the single-agent loop, its per-loop wiki, caps, the integrity gate and `ar.py` bookkeeping; used by `/autoresearch`
`neuroflow:wiki-protocol`	Knowledge-base protocol for the LLM-maintained wikis at three levels (flowie, project, hive) — ingest, query, lint and add workflows; used by `/wiki`, `/flowie --wiki-`, `/hive --wiki-` and `/autoresearch`

Domain skills¶

Skill	What it does
`neuroflow:review-neuro`	Rigorous eight-area peer review of a neuroscience manuscript — invoked by `/review`
`neuroflow:bids`	Brain Imaging Data Structure — folder hierarchy, entities, required files per modality, sidecars, derivatives, the validator, pybids and MNE-BIDS; loaded in the data phases
`neuroflow:humanizer`	Style editing on request — cuts filler, varies rhythm and matches the register while keeping every fact, number, citation and hedge; never a way to hide AI use, which is always disclosed
`neuroflow:pupil-labs-neon-realtime`	Real-time data streams (video, gaze, IMU, events) from Pupil Labs Neon eye-tracking glasses through the Real-time API

Integration skills¶

Skill	What it does
`neuroflow:setup-guide`	Configure integrations — Google Workspace, optional Miro and Anthropic-compatible LLM gateways — without ever asking for a secret in chat; used by `/setup`
`neuroflow:notebooklm`	Google NotebookLM through the `notebooklm-py` CLI — notebooks, sources (each upload confirmed first), generated artifacts and downloads

Phase skills¶

Each research phase has a corresponding skill that orients Claude's approach, suggests relevant domain tools, and provides workflow hints.

Skill	Associated command
`neuroflow:phase-git`	`/git`
`neuroflow:phase-ideation`	`/ideation`
`neuroflow:phase-preregistration`	`/preregistration`
`neuroflow:phase-grant-proposal`	`/grant-proposal`
`neuroflow:phase-finance`	`/finance`
`neuroflow:phase-experiment`	`/experiment`
`neuroflow:phase-tool-build`	`/tool-build`
`neuroflow:phase-tool-validate`	`/tool-validate`
`neuroflow:phase-data`	`/data`
`neuroflow:phase-data-preprocess`	`/data-preprocess`
`neuroflow:phase-data-analyze`	`/data-analyze`
`neuroflow:phase-paper`	`/paper`
`neuroflow:phase-review`	`/review`
`neuroflow:phase-poster`	`/poster`
`neuroflow:phase-notes`	`/notes`
`neuroflow:phase-write-report`	`/write-report`
`neuroflow:phase-slideshow`	`/slideshow`
`neuroflow:phase-brain-build`	`/brain-build`
`neuroflow:phase-brain-optimize`	`/brain-optimize`
`neuroflow:phase-brain-run`	`/brain-run`
`neuroflow:phase-quiz`	`/quiz`
`neuroflow:phase-fails`	`/fails`
`neuroflow:phase-output`	`/output`
`neuroflow:phase-pipeline`	`/pipeline`
`neuroflow:phase-search`	`/search`
`neuroflow:phase-flowie`	`/flowie`
`neuroflow:phase-hive`	`/hive`
`neuroflow:phase-meeting`	`/meeting`

Development skills¶

Skill	What it does
`neuroflow:neuroflow-develop`	Guide for developing and maintaining the neuroflow plugin
`neuroflow:skill-creator`	Guide for creating new neuroflow skills

For contributors¶

If you want to add a new skill:

Create a folder under skills/ with the skill name: skills/my-new-skill/
Add a SKILL.md file with the frontmatter and content
Reference the skill from the relevant command with Read the neuroflow:my-new-skill skill first

See the neuroflow:skill-creator skill for the full authoring guide.

Skills must not create their own folders in .neuroflow/

All skill output must be written to the active command's phase subfolder (e.g. .neuroflow/data-preprocess/). A skill named my-skill must not create .neuroflow/my-skill/. The sentinel-dev agent checks for this.