Intelligence

Use Unified API market intelligence to rank current setups and inspect one canonical symbol. Read the returned evidence; never treat a score as advice.

Files in this skill

File	What it covers
references/api.md	Unified scan/setup parameters and contract-shaped examples.
references/workflow.md	Scan → setup → backtest → deploy workflows.
references/buckets.md	Optional strategy lenses for interpreting evidence; these are not Unified response categories.
references/glossary.md	How to explain common evidence fields without assuming they are always present.

Route selection

• List/ranking question → GET /context/scan. Use bucket=all for an open-ended scan, or alts/hip3 when the user narrows the universe. Use category=all unless they explicitly request momentum, mean_reversion, breakout, or volume. top_n defaults to 10 and must not exceed 50.
• Single-symbol question → first resolve a canonical symbol with GET /context/markets, then call GET /context/setup/{symbol} with the URL-encoded canonical value.

Presenting results

1. State computed_at when the response includes it.
2. Present the returned symbol, category, and score without renaming them to an older bucket taxonomy.
3. Cite the exact evidence values returned for each conclusion.
4. Treat setup fields such as regime, levels, volatility, and funding as optional unless they appear in the response.
5. Before recommending capital, apply an appropriate strategy lens, run a backtest, and explain the principal risk.

Never manufacture an empty or unavailable scan from training data. Scores are deterministic market computations, not guarantees or a signal service.