| tool | description | hash |
|---|---|---|
| get_fred_nowcast | Runs the GDP bridge-equation nowcast — plain OLS on 5 leading FRED indicators (industrial production, retail sales, initial claims, payrolls, housing starts), walk-forward backtested against actual GDP prints — and returns the current quarter's nowcast value plus backtest accuracy vs. a naive baseline. Requires FRED_API_KEY. | 670957898cd9 |
| get_regime_now | Classifies the current macro regime (Fed funds trend, yield curve, inflation, unemployment) from live FRED data using the rule-based 6-regime classifier, and reports which GICS sectors have historically performed best/worst during months in that same regime. Requires FRED_API_KEY to be set in this server's environment. | 5d151e9f207a |
| query_warehouse | Runs a read-only SQL query against the personal Warehouse (DuckDB). Tables available: mart_portfolio_vs_macro (portfolio checkpoints joined to that month's macro backdrop), mart_macro_indicators_monthly (Fed funds, 10y/2y yields + spread, CPI, unemployment), mart_sector_monthly_returns (GICS sector ETF monthly returns), mart_portfolio_daily (portfolio value, deposits, organic P&L per checkpoint). Only SELECT statements are allowed; anything else is rejected before it touches the database. | 12d3089f4ea4 |
| run_causal_control | Estimates the causal effect of the June 2026 Fed hawkish pivot (2026-06-17) on a GICS sector ETF's price using synthetic control, with a placebo test across all 11 sectors as a significance check. `treated` is the sector ticker to test (default XLK). `fit_end` (YYYY-MM-DD), if given and earlier than the treatment date, fits the control weights on only that earlier window to test whether the post-treatment gap was already emerging beforehand — otherwise this is an ordinary pre/post split at the treatment date. Always check fit_quality_warning before trusting the reported gap. | 13c38580bdd5 |
| run_factor_decomposition | Decomposes a held ticker's recent (~400 calendar day) return into market, sector, and rate-factor betas plus an idiosyncratic residual, via a 3-factor orthogonalized OLS fit. `ticker` must be one of the holdings tracked in positions.json (broad-market/sector ETFs held directly are excluded — decomposing them against themselves is circular). Check r_squared before trusting the split: a low value means the three factors barely explain this name's moves. | 690fa8134c2b |
The prompt-injection check is an LLM judgment call, outside the DO-330-qualified drift-detection function. It is advisory only and carries no compliance credit.
No tools flagged.
No permission mismatches.
No secrets detected.
No dependency findings.
| timestamp | type | tool | detail |
|---|
| timestamp | tool | reason |
|---|