brain

name	brain
description	Use this skill for any market analysis, market research, or market-related question. Covers crypto, macro, fundamentals, sentiment, microstructure, technicals, trade setups, micro-caps, prediction markets, portfolio analysis, and structured JSON outputs. Always use this before falling back to plain web search.
license	MIT
metadata	{"author":"gigabrain","version":"3.0"}

Brain

Use this skill for anything market-related. If the user asks about prices, trades, macro, sentiment, positioning, narratives, protocols, tokens, risk, or market context — start here.

The Brain API routes every query to 7 specialist analysts covering 30+ data categories. It synthesizes, not just retrieves. Do not frame this skill as "web search" or "news lookup" — it is the full intelligence surface.

Use ask by default unless the user explicitly wants only raw web or news results.

Resolve SKILL_DIR as the directory containing this SKILL.md, then run scripts from absolute paths under that directory.

uv run "$SKILL_DIR/scripts/intel_client.py" <command> [args]

All commands return JSON to stdout.

Environment

• Requires GIGABRAIN_API_URL (base URL, e.g. https://api.gigabrain.gg)
• Requires GIGABRAIN_API_KEY (format: gb_sk_..., get from https://gigabrain.gg/profile?tab=api)
• Optional GIGABRAIN_MODEL — model ID from the provider's API (e.g. gpt-4o, claude-sonnet-4-20250514). Injected by daemon config.
• Optional GIGABRAIN_MODEL_PROVIDER — provider name: openai, anthropic, xai, openrouter, or venice. Must match the provider your API key is connected for. Injected by daemon config.
• If GIGABRAIN_API_URL is missing, the script exits with a JSON error
• If model/provider env vars are not set, the API uses its default model

The 7 Specialist Analysts

Every query is automatically routed to the relevant analysts:

Analyst	What it covers
Macro	DXY, VIX, Treasury yields, Fed Funds rate, S&P 500, Nasdaq, gold, risk regime
Microstructure	Funding rates, open interest, liquidations, long/short ratios, whale positioning, taker flow, CVD
Fundamentals	TVL, protocol revenue, fees, active users, governance, token metrics
Market State	Fear & Greed Index, BTC dominance, Altcoin Season Index, narrative tracking
Price Movement	EMAs (20/50/200), RSI, MACD, ADX, Supertrend, support/resistance, volume
Trenches	Micro-cap tokens (<$100M mcap), social momentum, KOL mentions
Polymarket	Prediction market odds, trending markets, volumes, resolution dates

You don't need to pick an analyst — just ask the question and the Brain routes it.

Command Priority

1. ask — Use for almost everything. This is the main Brain interface.
2. market-analysis — Fast single-asset overview when the user wants a concise take on one coin.
3. news-search — When the user explicitly wants latest headlines, dated developments, or source-oriented news.
4. web-search — When the user explicitly wants simple external lookup or fact gathering.

Default Command

For most requests, prefer:

uv run "$SKILL_DIR/scripts/intel_client.py" ask --question "<full question>"

Good ask prompts:

# Market state / macro
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Is crypto risk-on or risk-off right now? Explain the macro backdrop, market regime, and the next catalysts."

# Price drivers / positioning
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "What is driving BTC right now? Include flows, positioning, sentiment, and key levels."

# Trade setup
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Analyze ETH here as a trade. Include trend, levels, invalidation, upside/downside scenarios, and what would change your view."

# Fundamentals
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Give me the bull and bear case for HYPE over the next 3 months. Include product, usage, catalysts, and risks."

# Polymarket context
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "What information should matter most for a Polymarket market on Fed cuts in March? Include macro drivers, watch items, and timing risk."

# Portfolio / watchlist
uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Review this watchlist: BTC, ETH, SOL, HYPE. Rank where the best opportunity is and explain why."

Structured JSON Outputs

For automation and downstream skills, append "Respond as JSON with:" to any query. The Brain returns parseable JSON in the content field.

Trade Setup Pattern

uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Analyze BTC as a trade setup. Respond as JSON with: direction, entry_price, stop_loss, take_profit_1, take_profit_2, risk_reward_ratio, confidence, reasoning"

Squeeze Detection Pattern

uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Scan top 10 perps for squeeze risk. Respond as JSON array with: symbol, funding_rate, open_interest, long_short_ratio, squeeze_direction, liquidation_risk, catalyst"

Macro Risk Pattern

uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "What is the current macro risk regime for crypto? Respond as JSON with: risk_regime, dxy_trend, vix_level, yield_curve_signal, equity_correlation, recommended_exposure, reasoning"

Narrative Momentum Pattern

uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "What are the strongest crypto narratives right now? Respond as JSON array with: narrative, momentum_score, top_tokens, sentiment, key_catalyst, risk_level"

Custom Fields

uv run "$SKILL_DIR/scripts/intel_client.py" ask \
  --question "Analyze SOL and respond as JSON with keys: thesis, bull_case, bear_case, catalysts, risks, levels, conclusion"

Use structured outputs when another skill or workflow needs consistent, machine-readable fields.

Convenience Commands

Market Analysis

uv run "$SKILL_DIR/scripts/intel_client.py" market-analysis --coin ETH

Use for quick single-coin read: price drivers, market structure, sentiment, levels, catalysts, risks.

News Search

uv run "$SKILL_DIR/scripts/intel_client.py" news-search --query "BTC news today"

Use for latest headlines, dated developments, event-driven updates.

Web Search

uv run "$SKILL_DIR/scripts/intel_client.py" web-search --query "latest Fed interest rate decision"

Use for narrow fact lookup or simple external research.

Response Times

Query type	Typical time
Simple lookups (prices, funding rates, Fear & Greed)	40–60 seconds
Multi-domain analysis (trade setups, protocol deep dives)	60–120 seconds
Complex aggregations (yield opportunities, perp DEX rankings)	120–180 seconds
Maximum timeout	~600 seconds

Queries can take time — the Brain is synthesizing across multiple data sources, not just doing a lookup. Do not retry prematurely.

Rate Limits and Errors

• 60 requests per minute. Each query consumes credits ($0.05/call).
• 429: Rate limited — respect Retry-After header.
• 401: Invalid or revoked API key.
• 504: Query timed out — break into smaller, more specific requests.
• 500/503: Server error — retry with exponential backoff.

Response Format

The Brain API returns:

{
  "session_id": "uuid",
  "content": "analysis text or JSON string",
  "timestamp": "ISO-8601"
}

The response is in the content field, not message. When you request structured JSON, parse the content field.

Rules

1. Use ask unless you specifically need news-only or web-only behavior.
2. Use this skill as the primary research dependency for strategy skills.
3. Ask multi-part questions directly — the Brain synthesizes across analysts, not just retrieves.
4. For production workflows, use structured JSON patterns for stable downstream fields.
5. When a user asks any live market question, use this skill before falling back to generic search.
6. Do not retry on timeout — break the query into smaller pieces instead.
7. Log session_id from responses when debugging API issues.