@battlegrid/mcp-server 32.0.86 → 32.0.89

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -76,10 +76,10 @@ weak/developing/trending/extreme), `includeMovingAverages` (`dist_EMA20`, `MAali
76
76
 
77
77
  Platform `includePerpSpotFlow` (`perpSpotFlow` vocab: confirmed_bull / confirmed_bear /
78
78
  perp_led_fragile / spot_led_accumulation / neutral), `includeStructureZones`
79
- (`zones_htf_support_dist` signed % of price from the zone midpoint, positive above a support zone; `_age_h`), custom `Tape`
79
+ (`zones_htf_support_dist` signed % of price from the zone's near edge (its high), positive above a support zone and 0 inside it; `_age_bars`, whole bars of the zone timeframe), custom `Tape`
80
80
  (`BUY_PRESSURE value`→`buyPres` 0–1, `RVOL value`, `CLOSE_CHANGE value`). Conditions:
81
81
  `SPOT_ACCUM` (`perpSpotFlow is "spot_led_accumulation"`); `NEAR_SUPPORT`
82
- (`zones_htf_support_dist between 0 2` + `zones_htf_support_age_h gte 12`); `DIP_BID` verdict
82
+ (`zones_htf_support_dist between 0 2` + `zones_htf_support_age_bars gte 3`); `DIP_BID` verdict
83
83
  UP — refs plus an `N_OF(2)` quorum over `buyPres gte 0.55` / `RVOL20 gte 1.2` / `closeChg gt 0`
84
84
  (quorum beats a brittle ALL); `FRAGILE_POP` verdict DOWN (`perp_led_fragile` +
85
85
  `zones_htf_resist_dist between -2 0` + `RVOL20 gte 1.3`). Levels 0.5–1.2 ATR, RR 1.5, ATR floor
@@ -16,6 +16,7 @@ rather than events all live there. This document assumes them.
16
16
  - Column recipes
17
17
  - Cross detection
18
18
  - Recency and structure
19
+ - Candle at a swing and liquidity-pool sweeps
19
20
  - Board-relative rank recipes
20
21
  - Cross-venue and perp/spot basis
21
22
  - Crowd positioning
@@ -31,6 +32,8 @@ rather than events all live there. This document assumes them.
31
32
  | Momentum build-up | `{ "metric": "RSI14", "transformId": "trajectory", "timeframe": { "rel": "anchor" }, "window": 4 }` | `RSI14_t3…_now`, `RSI14_trend` |
32
33
  | Closed-bar-only trajectory | add `"bars": "closed"` to any trajectory | same, drawn from closed bars |
33
34
  | Extension from trend | `{ "metric": "SMA50", "transformId": "distance", "timeframe": { "rel": "anchor" } }` | `dist_SMA50` (signed %) |
35
+ | Touch of the 20 EMA within half an ATR | `{ "metric": "EMA20", "transformId": "levelInteraction", "timeframe": { "rel": "anchor" }, "touchAtr": 0.5 }` | `EMA20_lvl_a0.5` — a column carries a buffer or an ATR multiple, never both |
36
+ | Touch of VWAP within 10 bps | `{ "metric": "VWAP", "transformId": "levelInteraction", "timeframe": { "rel": "anchor" }, "buffer": 10 }` | `VWAP_lvl_b10` |
34
37
  | Funding paid over a day | `{ "metric": "FUNDING_RATE", "transformId": "aggregate", "timeframe": { "rel": "anchor" }, "window": 24 }` | `rate_mean24` |
35
38
  | Chop filter on the entry rung | `{ "metric": "CLOSE", "transformId": "efficiency", "timeframe": { "rel": "lower" }, "window": 4, "bars": "closed" }` | `close_ltf_er` (≥0.6 directional; 4 changes, the last 5 closes) |
36
39
  | One-bar volume concentration | `{ "metric": "VOLUME", "transformId": "maxShare", "timeframe": { "rel": "lower" }, "window": 4, "bars": "closed" }` | `volBase_ltf_maxShare` |
@@ -132,6 +135,109 @@ step down of the decided pivot, which is a lower high against the pivot before i
132
135
  "definition": { "kind": "clause", "column": { "sectionKey": null, "header": "fractalHi" }, "op": "decreased", "hold": { "atLeast": 2, "of": 8 } } }
133
136
  ```
134
137
 
138
+ ## Candle at a swing and liquidity-pool sweeps
139
+
140
+ A candle pattern AT a swing, and a sweep OF a liquidity pool, are compositions of shipped columns — no
141
+ one column reads either whole. Each recipe below is a whole condition; adapt its verdict and its `required`,
142
+ and keep its shape.
143
+
144
+ | Intent | Column | Header |
145
+ |---|---|---|
146
+ | Swing low age at a 21/21 pair | `{ "metric": "SWING_FRACTAL_LOW_BARS_AGO", "transformId": "value", "timeframe": { "rel": "anchor" }, "leftBars": 21, "rightBars": 21 }` | `fractalLoAgo_l21r21` |
147
+ | Swing high age at a 21/21 pair | `{ "metric": "SWING_FRACTAL_HIGH_BARS_AGO", "transformId": "value", "timeframe": { "rel": "anchor" }, "leftBars": 21, "rightBars": 21 }` | `fractalHiAgo_l21r21` |
148
+ | Pin bar of the bar 21 back | `{ "metric": "CANDLE_PIN", "transformId": "value", "timeframe": { "rel": "anchor" }, "offset": 21 }` | `pin_o21` |
149
+ | Engulfing of the bar 21 back | `{ "metric": "CANDLE_ENGULFING", "transformId": "value", "timeframe": { "rel": "anchor" }, "offset": 21 }` | `engulf_o21` |
150
+ | Buyside pool interaction | `{ "metric": "LIQUIDITY_POOL_ABOVE", "transformId": "levelInteraction", "timeframe": { "rel": "anchor" } }` | `poolAbove_lvl` |
151
+ | Sellside pool interaction | `{ "metric": "LIQUIDITY_POOL_BELOW", "transformId": "levelInteraction", "timeframe": { "rel": "anchor" } }` | `poolBelow_lvl` |
152
+ | Closed back under the buyside pool the last bar retired | `{ "metric": "LIQUIDITY_POOL_ABOVE_RECLAIMED", "transformId": "value", "timeframe": { "rel": "anchor" } }` | `poolAboveRecl` |
153
+ | Closed back over the sellside pool the last bar retired | `{ "metric": "LIQUIDITY_POOL_BELOW_RECLAIMED", "transformId": "value", "timeframe": { "rel": "anchor" } }` | `poolBelowRecl` |
154
+
155
+ **Candle at a swing** — LuxAlgo's *Swing Highs/Lows & Candle Patterns* at its length 21: a swing pivot of
156
+ 21 bars each side, and the candle pattern of the pivot bar itself, Pine's `[21]`. A pivot is decided on
157
+ the bar that completes its right side, where the swing age reads the pair's `rightBars`, so on that bar
158
+ the pivot bar is `rightBars` bars back and the pattern column reads it at that offset. **The offset
159
+ equals the swing pair's `rightBars` and changes with it**: a 10/5 pair reads `fractalLoAgo_l10r5 between 5 5`
160
+ beside `pin_o5`. The age never reads below `rightBars`, so `between 21 21` is "decided on this bar": the
161
+ condition reads TRUE once per pivot, on its deciding bar — where LuxAlgo draws the label back on the pivot
162
+ bar, the condition never paints back. The patterns are the `candlePatterns` rules (the skill body,
163
+ `## Bar geometry`), each judged against the ten bars before the pivot bar. LuxAlgo's script tests bullish
164
+ and bearish engulfing with one identical condition; `engulf` reads `bullish` for a white body over a black
165
+ one and `bearish` for the mirror, so that defect never arises here. An offset column and a column off the
166
+ 2/2 pair each hold Once.
167
+
168
+ A hammer at the swing low:
169
+
170
+ ```json
171
+ { "conditionKey": "HAMMER_AT_SWING_LOW", "name": "Hammer at a 21/21 swing low", "verdict": "UP", "required": false, "exitSide": null,
172
+ "definition": { "kind": "group", "op": "ALL", "hold": { "atLeast": 1, "of": 1 }, "members": [
173
+ { "kind": "clause", "column": { "sectionKey": null, "header": "fractalLoAgo_l21r21" }, "op": "between", "low": 21, "high": 21, "hold": { "atLeast": 1, "of": 1 } },
174
+ { "kind": "clause", "column": { "sectionKey": null, "header": "pin_o21" }, "op": "is", "label": "lower_wick", "hold": { "atLeast": 1, "of": 1 } } ] } }
175
+ ```
176
+
177
+ A shooting star at the swing high, the mirror:
178
+
179
+ ```json
180
+ { "conditionKey": "SHOOTING_STAR_AT_SWING_HIGH", "name": "Shooting star at a 21/21 swing high", "verdict": "DOWN", "required": false, "exitSide": null,
181
+ "definition": { "kind": "group", "op": "ALL", "hold": { "atLeast": 1, "of": 1 }, "members": [
182
+ { "kind": "clause", "column": { "sectionKey": null, "header": "fractalHiAgo_l21r21" }, "op": "between", "low": 21, "high": 21, "hold": { "atLeast": 1, "of": 1 } },
183
+ { "kind": "clause", "column": { "sectionKey": null, "header": "pin_o21" }, "op": "is", "label": "upper_wick", "hold": { "atLeast": 1, "of": 1 } } ] } }
184
+ ```
185
+
186
+ A bullish engulfing at the swing low:
187
+
188
+ ```json
189
+ { "conditionKey": "BULLISH_ENGULFING_AT_SWING_LOW", "name": "Bullish engulfing at a 21/21 swing low", "verdict": "UP", "required": false, "exitSide": null,
190
+ "definition": { "kind": "group", "op": "ALL", "hold": { "atLeast": 1, "of": 1 }, "members": [
191
+ { "kind": "clause", "column": { "sectionKey": null, "header": "fractalLoAgo_l21r21" }, "op": "between", "low": 21, "high": 21, "hold": { "atLeast": 1, "of": 1 } },
192
+ { "kind": "clause", "column": { "sectionKey": null, "header": "engulf_o21" }, "op": "is", "label": "bullish", "hold": { "atLeast": 1, "of": 1 } } ] } }
193
+ ```
194
+
195
+ A bearish engulfing at the swing high, the mirror:
196
+
197
+ ```json
198
+ { "conditionKey": "BEARISH_ENGULFING_AT_SWING_HIGH", "name": "Bearish engulfing at a 21/21 swing high", "verdict": "DOWN", "required": false, "exitSide": null,
199
+ "definition": { "kind": "group", "op": "ALL", "hold": { "atLeast": 1, "of": 1 }, "members": [
200
+ { "kind": "clause", "column": { "sectionKey": null, "header": "fractalHiAgo_l21r21" }, "op": "between", "low": 21, "high": 21, "hold": { "atLeast": 1, "of": 1 } },
201
+ { "kind": "clause", "column": { "sectionKey": null, "header": "engulf_o21" }, "op": "is", "label": "bearish", "hold": { "atLeast": 1, "of": 1 } } ] } }
202
+ ```
203
+
204
+ **Sweep of a liquidity pool** — two forms. On ONE candle, the bar's high traded through the nearest
205
+ buyside pool above price and its close came back under it: `levelInteraction` over `LIQUIDITY_POOL_ABOVE`
206
+ reads `reject_from_below` (sellside: `reject_from_above` on `LIQUIDITY_POOL_BELOW`). Over TWO candles, a
207
+ close beyond the pool retires it (`poolAboveOk` reads false on that bar) and the next bar closes back:
208
+ from that next bar `poolAbove` already reads the next pool up, so the second bar is read through the
209
+ retired pool instead — `LIQUIDITY_POOL_ABOVE_RETIRED` (`poolAboveRetired`) is the pool the previous bar's
210
+ close retired, on that one bar only, and `LIQUIDITY_POOL_ABOVE_RECLAIMED` (`poolAboveRecl`) is true when
211
+ this bar's close is back at or under it. The "same or next candle" sweep is an `ANY` over the two. The
212
+ retired pool is a level, so `poolAboveRetired_lvl is "reject_from_above"` reads a retest that held after
213
+ the breakout, and `dist_poolAboveRetired` how far the close sits from it.
214
+
215
+ ```json
216
+ { "conditionKey": "BUYSIDE_POOL_SWEPT", "name": "Swept the buyside pool and closed back under it", "verdict": "DOWN", "required": false, "exitSide": null,
217
+ "definition": { "kind": "clause", "column": { "sectionKey": null, "header": "poolAbove_lvl" }, "op": "is", "label": "reject_from_below", "hold": { "atLeast": 1, "of": 1 } } }
218
+ ```
219
+
220
+ ```json
221
+ { "conditionKey": "SELLSIDE_POOL_SWEPT", "name": "Swept the sellside pool and closed back over it", "verdict": "UP", "required": false, "exitSide": null,
222
+ "definition": { "kind": "clause", "column": { "sectionKey": null, "header": "poolBelow_lvl" }, "op": "is", "label": "reject_from_above", "hold": { "atLeast": 1, "of": 1 } } }
223
+ ```
224
+
225
+ The same or next candle, on either side:
226
+
227
+ ```json
228
+ { "conditionKey": "BUYSIDE_POOL_SWEPT_TWO_BAR", "name": "Swept the buyside pool and closed back under it, on the same or the next candle", "verdict": "DOWN", "required": false, "exitSide": null,
229
+ "definition": { "kind": "group", "op": "ANY", "hold": { "atLeast": 1, "of": 1 }, "members": [
230
+ { "kind": "clause", "column": { "sectionKey": null, "header": "poolAbove_lvl" }, "op": "is", "label": "reject_from_below", "hold": { "atLeast": 1, "of": 1 } },
231
+ { "kind": "clause", "column": { "sectionKey": null, "header": "poolAboveRecl" }, "op": "is", "label": "true", "hold": { "atLeast": 1, "of": 1 } } ] } }
232
+ ```
233
+
234
+ ```json
235
+ { "conditionKey": "SELLSIDE_POOL_SWEPT_TWO_BAR", "name": "Swept the sellside pool and closed back over it, on the same or the next candle", "verdict": "UP", "required": false, "exitSide": null,
236
+ "definition": { "kind": "group", "op": "ANY", "hold": { "atLeast": 1, "of": 1 }, "members": [
237
+ { "kind": "clause", "column": { "sectionKey": null, "header": "poolBelow_lvl" }, "op": "is", "label": "reject_from_above", "hold": { "atLeast": 1, "of": 1 } },
238
+ { "kind": "clause", "column": { "sectionKey": null, "header": "poolBelowRecl" }, "op": "is", "label": "true", "hold": { "atLeast": 1, "of": 1 } } ] } }
239
+ ```
240
+
135
241
  ## Board-relative rank recipes
136
242
 
137
243
  `rank` turns any column into an ordinal across the scanned board. Ordering semantics: `hi` most
@@ -13,6 +13,9 @@ engine) and names its substitutions where the catalog lacks a primitive.
13
13
  - Donchian / Turtle breakout
14
14
  - ICT / SMC (FVG + order blocks)
15
15
  - Pivot highs / lows, HH-HL structure and break of structure
16
+ - Any pivot length (`leftBars` / `rightBars`)
17
+ - Swing Highs/Lows & Candle Patterns [LuxAlgo]
18
+ - Liquidity pools and their sweeps
16
19
  - Chaikin Money Flow
17
20
  - Now native, formerly substituted
18
21
  - Statistics over a series' own history
@@ -63,13 +66,15 @@ rules about the columns themselves, not about porting — they live in the skill
63
66
  - **VWAP reversion** → `dist_VWAP` band + `dist_VWAP_rank_far lte 5` for board-relative
64
67
  stretch; required `NOT [ADX14_state in ["trending","extreme"]]` veto; scalp geometry +
65
68
  aggressive timeDecay (`VWAP` anchors daily at 00:00 UTC; `VWAP_WEEK`, `VWAP_MONTH`, `VWAP_TRADING_SESSION` and
66
- the rolling `VWAP_ROLLING_1H`/`4H`/`24H` are the other anchors, each with ±1/2/3σ bands `_UB<k>`/`_LB<k>` — a
67
- band-touch entry reads `CLOSE × spread(VWAP_LB2)` rather than a hand-computed σ).
69
+ the rolling `VWAP_ROLLING_1H`/`4H`/`24H` are the other anchors, each with ±1/2/3σ bands `_UB<k>`/`_LB<k>` — the
70
+ close's position against a band is `CLOSE × spread(VWAP_LB2)` rather than a hand-computed σ, and the bar's contact
71
+ with it is `VWAP_LB2 × levelInteraction`).
68
72
  - **Donchian / Turtle breakout** (`close > ta.highest(high, 20)[1]`) → `dist_donchianHiPrior20 crossesAbove 0`
69
73
  over `DONCHIAN_UPPER_PRIOR × distance` (the breakout bar; `dist_donchianHiPrior20 gt 0` holds on every
70
74
  close of the run) + `RVOL20 gte 1.5`; turtle exits = trend preset (trail from 1R, giveback 50). Mirror with
71
- `dist_donchianLoPrior20 crossesBelow 0`. The same-bar rail (`dist_donchianHi20`) includes the breakout bar's
72
- own high, so it can never confirm a close-based break — do not AND it in. `zone is "breakout high"` reads
75
+ `dist_donchianLoPrior20 crossesBelow 0`. The same-bar rail is refused under `distance`: it includes the
76
+ breakout bar's own high, so it can never confirm a close-based break. Its gap, `close_donchianHi20_spread`, is
77
+ never above 0 — do not AND it in. `zone is "breakout high"` reads
73
78
  the prior-bar rail: it is the label form of the `dist_donchianHiPrior20 gt 0` state, holds on every close of
74
79
  the run, and adds nothing to the crossing; `"breakdown low"` mirrors it.
75
80
  **Daily-breakout variant on any anchor:** pin the rail at 1d — `dist_donchianHiPrior20_1d crossesAbove 0`
@@ -92,29 +97,71 @@ rules about the columns themselves, not about porting — they live in the skill
92
97
  `DAY_OF_WEEK between 1 and 5`) is an ordinary condition; the glosses state the windows.
93
98
  - **ICT / SMC (FVG + order blocks)** → `STRUCT_ZONES` is the native zone engine:
94
99
  `zones_htf_support_type` (`bullish FVG`/`bullish order block`), `zones_htf_support_dist
95
- between 0 1.5`, `_age_h gte 12`, HTF bias required via `MAalign_htf`. A killzone is a session window
100
+ between 0 1.5`, `_age_bars gte 3`, HTF bias required via `MAalign_htf`. A killzone is a session window
96
101
  (`TRADING_SESSION`, `UTC_HOUR`) and a sweep of a previous session's extreme is `PRIOR_<S>_HIGH_SWEPT` /
97
- `_LOW_SWEPT`; displacement and event *sequencing* are shapes the
98
- grammar does not have — a clause compares one column with a literal or with its own previous
99
- completed bar, never with another column or an earlier event, so an ordered sequence of events
100
- cannot be stated at all. That is a grammar limit, not a missing metric; name it as
101
- one.
102
- - **Pivot highs / lows, HH-HL structure and break of structure** (`ta.pivothigh(2, 2)` /
103
- `ta.pivotlow(2, 2)`) → `SWING_FRACTAL_HIGH` / `SWING_FRACTAL_LOW` are the native swing levels and
102
+ `_LOW_SWEPT`. Event *sequencing* is an entry condition's root **In order** (`op: "SEQUENCE"`,
103
+ `within` bars — see `## Conditions` in the skill body): sweep → displacement → return to the zone is
104
+ `In order within 12 [ PRIOR_ASIA_LOW_SWEPT step, the displacement step, the zone-touch step ]`, each
105
+ step one bar's reading. What it cannot state is still a grammar limit, not a missing metric: an exit
106
+ that waits on an ordering, an ordering inside another, or a step that counts its own closes — name
107
+ it as one.
108
+ - **Pivot highs / lows, HH-HL structure and break of structure** (`ta.pivothigh(left, right)` /
109
+ `ta.pivotlow(left, right)`) → `SWING_FRACTAL_HIGH` / `SWING_FRACTAL_LOW` are the native swing levels and
104
110
  `SWING_FRACTAL_LABEL` the structure label (`HH`/`HL`/`LH`/`LL`/`none`) — see `## Swing structure` in
105
- the skill body. Pine's tie rule holds (an equal extreme on the left is allowed, on the right it is
106
- not), and each level is the stepped, non-repainting series — held from the bar that completes the
107
- pivot's right side, not plotted back on the pivot bar. Break of structure is
108
- `dist_fractalHi crossesAbove 0` / `dist_fractalLo crossesBelow 0`; "structure is bullish" is
109
- `fractalLabel in ["HH","HL"]`. The pivot each latest one is compared with is native too:
111
+ the skill body, which states the default pair, `ta.pivothigh(2, 2)`. Pine's tie rule holds (an equal
112
+ extreme on the left is allowed, on the right it is not), and each level is the stepped, non-repainting
113
+ series — held from the bar that completes the pivot's right side, not plotted back on the pivot bar.
114
+ Break of structure is `dist_fractalHi crossesAbove 0` / `dist_fractalLo crossesBelow 0`; "structure is
115
+ bullish" is `fractalLabel in ["HH","HL"]`. The pivot each latest one is compared with is native too:
110
116
  `PRIOR_SWING_FRACTAL_HIGH` / `PRIOR_SWING_FRACTAL_LOW` (`pFractalHi` / `pFractalLo`, Pine's
111
117
  `ta.valuewhen(not na(ph), ph, 1)`) are price levels held while the prior pivot is in reach, so
112
118
  `dist_pFractalHi crossesAbove 0` is a close through the swing high before the last one. A pivot's age is
113
119
  `SWING_FRACTAL_HIGH_BARS_AGO` / `SWING_FRACTAL_LOW_BARS_AGO` (`fractalHiAgo` / `fractalLoAgo`), counted
114
- from the pivot's own bar: 2 on the bar that completes it, so it reads Pine's
115
- `ta.barssince(not na(ph)) + 2`, and `fractalLoAgo lte 5` keeps a setup to a fresh swing low. Only two
116
- bars each side is native: a `ta.pivothigh(5, 5)` or longer script ports onto the two-bar swing, and the
117
- spec-lock question says so.
120
+ from the pivot's own bar: `rightBars` on the bar that completes it, so it reads Pine's
121
+ `ta.barssince(not na(ph)) + rightBars` (`+ 2` at the default), and `fractalLoAgo lte 5` keeps a setup to
122
+ a fresh swing low — the deciding bar and the three after it, which at 5/5 is `fractalLoAgo_l5r5 lte 8`.
123
+ - **Any pivot length ports exactly — `ta.pivothigh(5, 5)` is the swing column at 5/5, not a
124
+ substitution.** The seven swing metrics (the two levels, the two prior pivots, the two ages and the
125
+ label) declare `leftBars` / `rightBars`: default 2/2, minimum 1, `bound: 'depth'`. Stage Pine's two
126
+ lengths on the swing column beside its transform knobs, and leave the pair off for 2/2 — an explicit
127
+ 2/2 is the same column. This column is the distance to `ta.pivothigh(5, 5)`'s held level, with the same
128
+ tie rule and held step:
129
+ `{ "metric": "SWING_FRACTAL_HIGH", "transformId": "distance", "timeframe": { "rel": "anchor" }, "leftBars": 5, "rightBars": 5 }`.
130
+ Off 2/2 the header carries `_l<left>r<right>` after the code: `dist_fractalHi_l5r5`, `fractalLabel_l5r5`,
131
+ `fractalLoAgo_l5r5` (`fractalHi_l5r5_1h` pinned to 1h).
132
+ The pair reads `left + right + 1` bars to decide one pivot and must fit the readable depth; a wider pair
133
+ is refused with `REPORT_COLUMN_DEPTH_EXCEEDED` on `leftBars`, and each side's served `maximum` is the
134
+ bound (`strategy-authoring`'s knob rules). Off 2/2 the server computes the pivots over the column's own
135
+ bars, so two refusals follow:
136
+ - `rank`, `relativeTo` and the depth-bounded windows (the averages and the statistics) read the stored
137
+ 2/2 and refuse an off-default pair — stage a pair only on a transform whose entry serves
138
+ `metricParametersAccepted: true`;
139
+ - a condition over the column admits one bar: a hold over more than one close is refused
140
+ (`CONDITION_HOLD_ILLEGAL`), so is an **In order** step over it (`CONDITION_SEQUENCE_WINDOW_INVALID`),
141
+ and every prior-bar operator (`crossesAbove`, `crossesBelow`, `increased`, `decreased`, `enters`,
142
+ `exits`) is refused with `CONDITION_OPERATOR_UNSUPPORTED`, leaving `lt`, `lte`, `gte`, `gt`, `between`,
143
+ `is` and `in`.
144
+
145
+ So at 5/5 the levels, the label, the prior pivots and the ages port as states
146
+ (`fractalLabel_l5r5 in ["HH","HL"]`, `fractalLoAgo_l5r5 lte 8`), and the one-bar break the default pair
147
+ reads as `dist_fractalHi crossesAbove 0` reads at 5/5 as the state `dist_fractalHi_l5r5 gt 0`: the close
148
+ stands above the last 5/5 swing high, from the break until the close falls back or the level steps to
149
+ the next swing. That edge-for-state reading is what the spec-lock question names; the pivot length
150
+ itself is the script's own and is never named as a substitution.
151
+ - **Swing Highs/Lows & Candle Patterns [LuxAlgo]** (length 21, the pattern read at Pine's `[21]`, the
152
+ pivot bar) → the swing age at the 21/21 pair pinned to 21 beside the candle pattern at offset 21:
153
+ `fractalLoAgo_l21r21 between 21 21` AND `pin_o21 is "lower_wick"` is the hammer at a swing low, and
154
+ `engulf_o21` the engulfing. The offset equals the pair's `rightBars` and changes with it. The condition
155
+ reads TRUE on the bar that decides the pivot, where the script draws its label back on the pivot bar;
156
+ the patterns are the `candlePatterns` rules, so the script's identical bullish and bearish engulfing
157
+ test never arises. Whole conditions: `recipes.md`, *Candle at a swing and liquidity-pool sweeps*.
158
+ - **Liquidity pools and their sweeps** (LuxAlgo Liquidity Swings' close retirement over SMC's equal
159
+ pivots) → `LIQUIDITY_POOL_ABOVE` / `LIQUIDITY_POOL_BELOW` (`poolAbove` / `poolBelow`) are native. A
160
+ one-candle sweep is `poolAbove_lvl is "reject_from_below"`, mirrored with `poolBelow_lvl is
161
+ "reject_from_above"`. A sweep over two bars — the outbreak's close retires the pool, the next close
162
+ comes back — is `poolAboveRecl is "true"` (`poolBelowRecl`), and the retired pool itself is
163
+ `poolAboveRetired` / `poolBelowRetired`, a level a retest reads with `levelInteraction`. Whole
164
+ conditions: `recipes.md`, *Candle at a swing and liquidity-pool sweeps*.
118
165
  - **Chaikin Money Flow** (TradingView's built-in "Chaikin Money Flow" script, length 20; Pine has no
119
166
  `ta.cmf`) → `CMF20` is native: each bar's volume weighted by where it closed in its range, summed over
120
167
  20 closed bars and divided by their volume, −1 to +1. `CMF20 gt 0` is the accumulation state and
@@ -126,7 +173,8 @@ rules about the columns themselves, not about porting — they live in the skill
126
173
  and signals (`MACD_LINE`/`MACD_SIGNAL`, `PPO_LINE`/`PPO_SIGNAL` — the line's zero cross is
127
174
  `MACD_LINE × crossDetect`, and the signal-line cross is `MACD_HIST × crossDetect`, the histogram being
128
175
  line minus signal), the 15-bar rate of change (`ROC15`), the Donchian break memory
129
- (`DONCHIAN_BROKEN_HI`/`DONCHIAN_BROKEN_LO`), daily pivots of
176
+ (`DONCHIAN_BROKEN_HI`/`DONCHIAN_BROKEN_LO`, and as of the bar before — Pine's `brk[1]` —
177
+ `DONCHIAN_BROKEN_HI_PRIOR`/`DONCHIAN_BROKEN_LO_PRIOR`, the level a retest or failed-breakout clause reads), daily pivots of
130
178
  all six Pivot Points Standard types (Traditional `PIVOT_P`/`PIVOT_R1`–`R5`/`PIVOT_S1`–`S5`, then
131
179
  `PIVOT_FIB_*`, `PIVOT_CLASSIC_*`, `PIVOT_CAM_*`, `PIVOT_WOODIE_*`, `PIVOT_DM_*` — a level one type shares
132
180
  with Traditional is read from the Traditional key) and the same six types on the previous ISO week,
@@ -154,6 +202,12 @@ rules about the columns themselves, not about porting — they live in the skill
154
202
  - `ta.rsi(x, n)` and `ta.cci(x, n)` are `x × rsi` and `x × cci` at window `n` (`hlc3_rsi21`, `close_cci50`),
155
203
  each null over a window whose values are all equal; the stored `RSI14`, `RSI7`, `RSI2` and `CCI20` (on HLC3)
156
204
  are the exact source at their own length, and `EMA50 × rsi` is the RSI of the stored 50-bar average;
205
+ - one of those over another is a chain, the inner's length on `window` and the outer's on `chainedWindow`:
206
+ `ta.rsi(ta.ema(x, n), m)` is `x × ema(n) × rsi(m)` (`close_ema200_rsi14`), `ta.cci(ta.sma(x, n), m)` is
207
+ `x × aggregate(n) × cci(m)`, and the built-in RSI's smoothing line `ta.sma(ta.rsi(x, n), m)` is
208
+ `x × rsi(n) × aggregate(m)`. The outer may be any average or statistic; the chained outer's catalog entry names
209
+ the key it bounds (`windowParameter`) and serves `innerWindowMaximum` for the inner. Under an average or an RSI
210
+ the chain is the windowed composite, within a few RSI points of a full-history chart on a slow average;
157
211
  - `ta.percentrank(x, n)` is `x × pctl` at window `n` — the `n` bars BEFORE the newest, ties at or below;
158
212
  - `ta.highest(x, n)` / `ta.lowest(x, n)` are `x × rollingMax` / `x × rollingMin` (`high_max20`
159
213
  is `DONCHIAN_UPPER`'s same-bar rail);
@@ -39,9 +39,15 @@ Not "what were its returns" — whether it did *the thing it was told to do*.
39
39
  `signalSource`, and quote its KPIs as served — win rate, net P&L, best and worst trade, entry
40
40
  rate, average conviction and risk/reward, durations — recomputing none. `get_agent_performance`
41
41
  measures from the budget baseline, not over a period. A `{ profile: null }` result is
42
- **UNDETERMINED** for that agent: the read serves the player's own private agents only. The fixed
43
- Decision Quality window and its regime, hour, weekday and coin slices are
44
- `get_agent_review_analytics`'s, not this read's.
42
+ **UNDETERMINED** for that agent: the read serves the player's own private agents only. The regime,
43
+ hour, weekday and coin slices of the decisions made over a period are
44
+ `get_agent_review_slices`'s, not this read's.
45
+ - `get_agent_conditions` when the question is whether the mandate's conditions are doing their job: each
46
+ condition the agent's decisions cited over the last 90 days (the served `window`), by verdict against what
47
+ those decisions became — one row per strategy revision, because a revision may redefine a condition, so never
48
+ add two rows of one condition together. `get_agent_entry_discipline` answers what waiting for confirmation cost
49
+ over the same window. For every agent on one strategy, pass its `strategyId` to `get_conditions_breakdown` or
50
+ `get_entry_discipline_breakdown`.
45
51
  - `get_trade_outcome_by_decision` (a trade row's `decisionId`) / `get_trade_chart` when a specific trade
46
52
  needs explaining: the outcome carries the fills, fees, slippage, leverage, duration and exit label. A
47
53
  measurable trade's `tradeExcursion` reads in R: `mfeR` and `maeR` (the best and worst it reached),
@@ -76,7 +82,10 @@ the platform's recorded forward marks, never from today's price.
76
82
 
77
83
  - Select the decisions with `list_signal_logs`, then read each one's `get_signal_debrief` (the row's
78
84
  `agentId`, and its `id` as `logId`). When the question is the agent-wide pattern rather than named
79
- decisions, read `get_agent_decision_quality`.
85
+ decisions, read `get_agent_decision_quality`: one row per model the agent's decisions ran on under
86
+ `quality.rows`, the `isCurrentModel` row being the model it runs now — name a row's model beside every
87
+ figure quoted from it, and never pool two rows. For your account, a team or a strategy cohort, read
88
+ `get_decision_quality_breakdown`.
80
89
  - **Never mark a past decision against a current price.** A market-context read, a live mark or any
81
90
  price read now answers a different question. The debrief's marks are anchored at the decision
82
91
  (`decisionAnchor`) and, for a closed trade, at the exit (`exitAnchor`).