@battlegrid/mcp-server 31.2.1 → 31.2.3

package/dist/index.d.ts

@@ -49,7 +49,7 @@ import { type Implementation, type Prompt, type Resource } from '@modelcontextpr
  * being asked. Move it for a change to THIS package — a proxy fix, a dependency bump, a docs
  * correction. Never move it to track the server.
  */
-export declare const PACKAGE_VERSION = "31.2.1";
+export declare const PACKAGE_VERSION = "31.2.3";
 export declare const DEFAULT_URL = "https://mcp.battlegrid.trade/mcp";
 export interface EnvConfig {
     apiKeys: string[];

package/dist/index.js

@@ -52,7 +52,7 @@ import { ListToolsRequestSchema, CallToolRequestSchema, ListPromptsRequestSchema
  * being asked. Move it for a change to THIS package — a proxy fix, a dependency bump, a docs
  * correction. Never move it to track the server.
  */
-export const PACKAGE_VERSION = '31.2.1';
+export const PACKAGE_VERSION = '31.2.3';
 export const DEFAULT_URL = 'https://mcp.battlegrid.trade/mcp';
 const MAX_RETRIES = 3;
 const RETRY_DELAYS_MS = [2000, 4000, 8000];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@battlegrid/mcp-server",
-  "version": "31.2.1",
+  "version": "31.2.3",
   "description": "BattleGrid MCP server — play crypto prediction games from AI agents",
   "type": "module",
   "main": "dist/index.js",

package/skills/battlegrid-strategy-studio/references/recipes.md

@@ -34,11 +34,14 @@ ladder successor — 1d for a 4h anchor), pinned `{ "abs": "<tf>" }` → `_<tf>`
 Pins bind to discovery's `rankedTimeframes` (a superset of the anchor set), are fixed across
 anchor retunes, and `offset` does not change the header — one offset per `metric × timeframe`
 per section. `offset` is a `value`-transform parameter only: `distance` rejects it (typed
-error), so a level is always measured as it stands now — the literal *previous-day* high/low
-(PDH/PDL) is not addressable, and the daily swing levels (`zone_1d`, `dist_swingHi_1d`) are
-the named nearest level for daily-breakout intents. Conditions compare a column against a
-literal, never against another column. Rank ordering semantics: `hi` most positive, `lo` most
-negative, `far`/`near` by magnitude (offered only where sign matters).
+error), so a level is always measured as it stands now. That is a statement about `offset`, not
+about prior-day levels: `PDH`/`PDL` are published structure metrics at the daily anchor and take
+`distance` and `crossDetect`, so the literal previous-session extreme is authored as
+`dist_PDH_1d` rather than reconstructed from an offset. The daily *swing* levels (`zone_1d`,
+`dist_swingHi_1d`) are the tracked pivot — a different level, not a substitute. Conditions
+compare a column against a literal, never against another column. Rank ordering semantics:
+`hi` most positive, `lo` most negative, `far`/`near` by magnitude (offered only where sign
+matters).
 
 ## Condition patterns
 

package/skills/battlegrid-strategy-studio/references/tradingview-ports.md

@@ -37,12 +37,12 @@ closes while stops, trailing, and time decay keep managing the position intraday
 
 | TradingView script | Verdict | Port anchor |
 |---|---|---|
-| Squeeze Momentum [LazyBear] / TTM Squeeze | **Direct port** — Keltner is native | `BB_UPPER spread KC_UPPER`, `BB_LOWER spread KC_LOWER`, `bollinger_squeeze` |
+| Squeeze Momentum [LazyBear] / TTM Squeeze | **Direct port** — `KC_SQUEEZE` reads the state itself | `KC_SQUEEZE`, `bollinger_squeeze` |
 | Supertrend / UT Bot Alerts | **Direct port** for the state; flip still substituted | `ST_LINE`, `ST_DIR`, `EMA5_13` for the flip, position-management trailing |
 | Chandelier Exit | Direct port of the *mechanism* | trailing from entry (`trailingTriggerR: 0`) |
 | MACD + 200 MA filter | Direct process port | `MACD_cross`, `dist_SMA200`, `macd_bull_cross` |
 | Golden / Death Cross | Direct process port | `SMA50_SMA200_spread` + `_trend` |
-| RSI-2 (Connors) | Named substitute (no RSI-2; RSI7 is the closest series) | `RSI7`, `dist_SMA200`, time decay |
+| RSI-2 (Connors) | **Direct port** — `RSI2` is native | `RSI2`, `dist_SMA200`, time decay |
 | VWAP reversion | Direct process port | `dist_VWAP`, `dist_VWAP_rank_far` |
 | Donchian / Turtle breakout | Process port (N-bar channel → swing structure) | `zone`, `dist_swingHi`, `sr_resistance_break` |
 | ICT/SMC: FVG + Order Blocks | Direct process port of the zone logic | `STRUCT_ZONES` columns + `structure_*` signals |
@@ -145,10 +145,11 @@ the book out of single-name traps. Rules: `ma_sma200_above` 3 required, `ma_ema_
 **Source process.** In an uptrend (close > 200 SMA), buy panic dips (RSI(2) < 10); exit fast
 (cross of the 5-period average / a few bars).
 
-**Port — named substitute.** The catalog carries `RSI14` and `RSI7` only, so there is no `RSI2`;
-`RSI7` is the fastest RSI series
-and `lte 10` on it is a rarer, deeper panic — state that trade-off rather than hiding it.
-`DIP_BUY` verdict UP = `dist_SMA200 gt 0` (ref a required `ABOVE_200`) AND `RSI7 lte 10`.
+**Port — direct.** `RSI2` is native as of contract 49.1, so this ports exactly and no substitution
+note is owed. `DIP_BUY` verdict UP = `dist_SMA200 gt 0` (ref a required `ABOVE_200`) AND
+`RSI2 lte 10`. Keep the literal `lte 10` as the gate: `RSI2 × classifyZone` exists and reads on the
+same Connors 10/90 bands, but a zone label is a fixed reading while the literal is a threshold the
+author can see and tune.
 Rules: `rsi_oversold` 3 required `{"threshold": 25}` (RSI14 gate tuned toward the fast-dip
 regime — read the signal schema first) · `ma_sma200_above` 2 required · `bollinger_lower_touch`
 1. The source's fast exit is time, not price: `timeDecayEnabled: true`, grace 180, interval
@@ -181,10 +182,11 @@ in rules (its own params define the break). Turtle exits are the trend preset: s
 ATR, RR 2+, trail from 1R with giveback 50, no time decay. Mirror with `"breakdown low"` /
 `dist_swingLo` / `sr_support_break` for shorts. **Daily variant on any anchor:** pin the same
 structure at 1d — `zone_1d is "breakout high"`, `dist_swingHi_1d gte 0` (both validated) —
-which is the studio's daily-breakout condition. Note the substitution: the level is the
-tracked *daily swing* pivot, not the literal previous-day high — `distance` takes no `offset`
-and clauses never compare two columns, so PDH/PDL as exact levels are not addressable (see
-the not-expressible table).
+which is the studio's daily-breakout condition. That pins the tracked *daily swing* pivot. For
+the literal previous-session extreme instead, author `PDH`/`PDL` directly — both are native
+structure metrics at the daily anchor, and both take `distance` and `crossDetect`, so
+`dist_PDH_1d gte 0` or a `crossDetect` on `PDH` expresses the prior-day break exactly. Pick by
+which level the source actually means; neither is a substitution for the other.
 
 ## 9 · ICT / Smart Money Concepts: FVG + Order Blocks — 15m/1h
 
@@ -215,6 +217,19 @@ Williams %R (`WILLR14`), Stochastic RSI (`STOCH_RSI14`), Bollinger boundaries
 (`BB_UPPER`/`BB_LOWER`), ADX components (`DI_PLUS`/`DI_MINUS`). All arrived with contract 46.1 —
 every one of them was listed as inexpressible in this file before it.
 
+Contract 49.1 did it again with seven more: Connors RSI-2 (`RSI2`, and `RSI2 × classifyZone` on its
+own 10/90 bands), the TTM squeeze state (`KC_SQUEEZE`), the 9/21/50 EMAs (`EMA9`/`EMA21`/`EMA50`)
+and the prior-day levels (`PDH`/`PDL`). Two of those were standing absence claims in this file:
+RSI-2 was called a named substitute, and the prior-day levels were said to need a grammar form no
+metric could supply — true of the `distance` offset it cited, and never a reason a PUBLISHED level
+could not exist. Author the prior-day levels at `1d` and nowhere else: a level computed from the
+prior daily bar has no reading on a 15-minute one, and construction refuses it rather than
+resolving empty.
+
+Re-read the absence section against the vendored digest before trusting any row in it. That check
+is `skill-contract.test.ts`, and it is the only thing that catches a claim which raises no error
+and merely tells a paying author their strategy cannot be built.
+
 ## Not expressible — the catalog keys this needs
 
 The one section where a claim that the catalog LACKS something may live, and every row names the key
@@ -222,14 +237,14 @@ it denies so the claim stays checkable. A claim about the grammar's SHAPE (a cla
 column against a literal; `distance` rejects an `offset`) is permanent and belongs in prose above.
 A claim that a metric is absent belongs here, or nowhere.
 
+Only a bare backticked key counts as the claim — write the neighbour as a column code or an
+expression, never as a lone key, or the row denies a metric that is in fact served.
+
 | Script / primitive | Key the catalog would need | Nearest expressible neighbour |
 |---|---|---|
-| RSI-2 (Connors) | `RSI2` | `RSI7 lte 10` as a named fast-dip substitute |
-
-**PDH/PDL** stays inexpressible for a different reason, and the distinction matters: `distance`
-rejects an `offset` and a clause never compares two columns, so the *grammar* has no form for it.
-No metric would fix that. Use the daily swing structure (`zone_1d`, `dist_swingHi_1d`) and name the
-substitution.
+| Slow-baseline trend filters (100-period) | `SMA100` | `dist_SMA50` or `dist_SMA200`, whichever the source's horizon is nearer |
+| 200-period EXPONENTIAL baseline | `EMA200` | `dist_SMA200`. Deferred on measurement, not preference: at the platform's warmup floor an exponential 200 still carries 12.4% of its seed bar while a simple 200 is exact over the same window |
+| Prior-WEEK high/low | `PWH` / `PWL` | the prior-DAY levels at the daily anchor, or `zone_1w` structure |
 
 Never compile a "port" of these under the source's name without the substitution note — a
 player who asked for WaveTrend and silently got Stochastic has no way to learn otherwise.