--- name: regime-check description: Review the cross-asset backdrop using Slatemark's composite read, the cross-asset panel, and relevant follow-ups. --- Use only tools available on the user's plan. Read saved context before asking for missing inputs. Name unavailable, stale, partial, or suppressed evidence and continue independent authorized reads. An auth failure or missing broker tool does not establish a manual journal workflow. Never infer data that was not returned. Cite sources, dates, and coverage. On an authorization or rate-limit failure, stop that affected path; do not retry or use component calls to bypass the denial or throttle. Run a Regime Check using the connected Slatemark tools. It combines two independent reads; say which facts come from which. - `analyze_market_regime` is Slatemark's fixed composite: the rate complex, dollar trend, credit ratio, volatility, and cross-asset correlations. - `analyze_market_data_cross_asset_panel` reads a panel of instruments, ratios, and the fed funds futures path: the user's saved panel, or Slatemark's default when none is saved or a saved panel cannot be read. Call it without `panels` so a saved panel applies; pass `panels` only for instruments the user names in this request. Both use delayed public market data and need no broker link. If one read fails or is not on the user's plan, name it and continue with the other. Read the panel in this order: 1. **Coverage.** State `panel_source`. `default` is Slatemark's default set, not a panel the user chose; users choose their own under Cross-asset panel on their Slatemark Profile page. `saved_panel: "unavailable"` means a saved panel could not be read. Give each panel's `bar_date`, flag `latest_bar_may_be_partial`, and name every `unavailable` symbol and every panel carrying `error`. Leave a missing panel missing. 2. **State.** Describe each panel from its returned numbers only. A ratio panel compares two prices: a rising ratio means the numerator outperformed the denominator. It is not a price difference. 3. **Relationships.** Report `correlation_shifts` with both windows, the leading `largest_moves` by `scaled_move`, and where related panels agree or disagree, such as the yen and franc pairs. Correlations leave out `alignment.unshared_dates`. 4. **Fed funds path.** Report the `anchor` and each solved meeting's implied rate and change as returned. Where `gap` stops the path, name its meeting and `missing_contracts`; never extend the path past a gap or rebuild it from a continuous `/ZQ` chart. Do not restate the path as a forecast or as the probability of a decision. `effr` is a reference value; the path does not depend on it. A block carrying `error` has no path. 5. **The user's lens.** When the user has said what they watch a panel for, in this conversation or in saved context, attribute that reading to them. It is their view, not Slatemark's, and it does not change the returned numbers. Read the composite's components, windows, and coverage the same way. Use individual tools only for relevant gaps or a requested deeper comparison, such as real yields or financial conditions. Distinguish the computed facts from your inference, weigh conflicting components and panels, and describe what evidence would change the assessment. Do not present the read as Slatemark's conclusion, and do not treat it as a prerequisite for every single-name factual question or as a signal to act.