neuroskill-skills

A collection of NeuroSkill EXG skills for NeuroLoop™️ — a biometric AI companion powered by a real-time EXG device (Muse or OpenBCI). Skills are loaded contextually: the harness injects the relevant skill file into the system prompt when the user's message matches the skill's domain.

Skills

EXG Data & API Skills

| Skill | Description | |---|---| | neuroskill-transport | WebSocket & HTTP transport layer, port discovery, Quick Start, output modes (--json / --full / default), and every global CLI flag. | | neuroskill-status | status command — full system snapshot: device state, signal quality, EXG scores, band powers, ratios, embeddings, labels, 48 h sleep summary, and recording history. | | neuroskill-sessions | session and sessions commands — per-session metric breakdowns with first-half → second-half trend arrows, full session listing across all days, and Unix timestamp helpers. | | neuroskill-search | search and compare commands — ANN search for neurally similar EXG moments across all history, and A/B session comparison with metric deltas, trend directions, and UMAP enqueuing. | | neuroskill-sleep | sleep and umap commands — EXG-based sleep stage classification (Wake / N1 / N2 / N3 / REM) with efficiency, onset latency, bout analysis; and GPU-accelerated 3D UMAP projection of session embeddings for spatial state comparison. | | neuroskill-labels | label, search-labels, and interactive commands — creating timestamped EXG text annotations, semantic vector search over labels (HNSW), and a cross-modal 4-layer graph search combining text similarity, EXG similarity, and temporal label proximity. Supports Graphviz DOT export. | | neuroskill-streaming | say, listen, notify, calibrate, calibrations, timer, and raw commands — on-device TTS speech, real-time WebSocket event streaming (EXG, PPG, IMU, scores, labels), OS notifications, calibration profile management, focus timer, and raw JSON passthrough for unlisted commands. | | neuroskill-data-reference | Complete metric field reference — EXG band powers, ratios & indices (FAA, TBR, BAR, TAR, APF, SEF95, coherence, …), core scores (focus, relaxation, engagement, meditation, mood, cognitive load, drowsiness), complexity measures (Hjorth, permutation entropy, Higuchi FD, DFA, sample entropy, PAC), PPG/HRV fields, motion & artifact markers, sleep stage codes, neurological correlate indices, and consciousness metrics. | | neuroskill-recipes | Use-case recipes and scripting patterns — shell snippets for focus monitoring, stress tracking, sleep quality analysis, cognitive load queries, meditation tracking, cross-modal graph search, A/B session comparison, time-range queries, and automation with cron / Python / Node.js / HTTP. |

Protocol & Intervention Skills

| Skill | Description | |---|---| | neuroskill-protocols | Full guided-protocol repertoire — 70+ mind-body practices matched to EXG metric signals, organised across: attention & focus, stress & autonomic regulation, emotional regulation & mood, relaxation & alpha promotion, sleep & circadian, body & somatic, consciousness & integration, cognitive performance & motivation, headache & migraine, energy & alertness, hemispheric balance & breathing, deep relaxation, recovery & rest, deep meditation, extended emotional processing, autonomic & vagal, motor & embodiment, neck & cervical relief, eye exercises & visual recovery, morning routines, workout & gym, hydration, bathroom & movement breaks, music, social-media & digital-addiction, and dietary protocols. |

Install

# From npm (recommended)
npx skills add NeuroSkill-com/skill

# From GitHub
npx skills add NeuroSkill-com/skill

# From a local clone
npx skills add ./skill

# Install a single skill
npx skills add NeuroSkill-com/skill --skill neuroskill-status

# List all available skills without installing
npx skills add NeuroSkill-com/skill --list

How Contextual Loading Works

On every user message the harness:

1. Runs npx neuroskill status and injects the live EXG snapshot into the system prompt.
2. Detects domain signals in the user's prompt (stress, sleep, focus, protocols, …).
3. Runs the relevant neuroskill commands in parallel (neurological, session, search-labels, …).
4. If protocol intent is detected, reads skills/neuroskill-protocols/SKILL.md and injects the full protocol repertoire for that turn.
5. Injects the skill index (SKILL.md) so the model always knows every capability available.

Tools Available to the Agent

| Tool | Purpose | |---|---| | neuroskill_run | Run any neuroskill EXG command and return its output. | | neuroskill_label | Create a timestamped EXG annotation for the current moment (supports --context and --at flags). | | run_protocol | Execute a multi-step guided protocol with OS notifications, per-step timing, and EXG labelling. | | prewarm | Kick off a background npx neuroskill compare so results are ready when needed. | | memory_read | Read the agent's persistent memory file. | | memory_write | Write or append to the agent's persistent memory file. | | web_fetch | Fetch a URL and return its content. | | web_search | Search the web and return results. |

Citing

If you use this library in academic work, please cite it as:

@software{kosmyna2026openbci,
  author    = {Nataliya Kosmyna and Eugene Hauptmann},
  title     = {{neuroskill}: Skills to model Human State of Mind using NeuroSkill™️},
  year      = {2026},
  version   = {0.0.1},
  url       = {https://github.com/NeuroSkill-com/skill},
  license   = {AI100},
}

License

AI100