Market Snapshot

LLMs are bad at indicator arithmetic. This skill runs it in code.

When to Use

• Any time a decision depends on an indicator value or a level.
• Checking whether a strategy's mechanical trigger fired on the last closed bar.
• Getting a reference stop distance (0.5 × ATR(20), the stop the original bot trained on).

How It Works

In an autonomous cycle the bar briefing (bar_briefing, the harness-tools
skill) already holds this bar's values: m3 (EMAs, ADX, SuperTrend, Keltner,
ATR, VWAP, participation, the skip-rule counters in context, swings and
liquidity, lastBreak, sweeps, htfLevels, the nearest levels in points
and R), m15 and h1 (the higher frames). Use it, and compute only what it
doesn't hold.

1. Get closed bars, oldest first, in a file. Autonomous: the runner's bars
   file (the prompt names it). Otherwise the harness tool bars_fetch
   ({"symbol": "MNQ", "timeframe": 3}: the runner's default, 5000 bars)
   writes <FTH_HOME>/bars-scratch/MNQ-3m.json (the path it returns) and says when its last bar closed (see the
   harness-tools skill).
   Don't paste a get_bars reply into a file: thousands of bars are far more
   than a tool reply can carry. The default is enough for EMA(200), prior-day
   levels, and the multi-timeframe read up to the 4-hour trend (its EMA50
   vote needs about 4000 3-minute bars). The windowed pieces (cisd_ote,
   the regime) read their own trailing 500 bars, as their source did.
   3-minute bars match the strategy parameters (1-minute for the flow
   strategies).

2. (The file is the broker's get_bars JSON; a CSV file works too. The
   tools read only JSON and CSV.)

3. Call the harness tool market_snapshot with the file:

   market_snapshot {"bars": "/home/u/.futures-trading-harness/bars-scratch/MNQ-3m.json"}

   Override parameters with params, e.g. {"orbMinutes": 30, "emaSlow": 50},
   or the volume profile's: vpRowSize 0.25 (one-tick rows on the price grid,
   so levels are tick prices), vpRows, vpValueArea, vpLookback
   (bars in the rolling profile). Use the same settings as the strategy you
   are judging (its params).

4. Read the JSON:

   • trend: emaFast(9), emaSlow(20), ema50, ema200, adx(14) and its 5-bar slope,
     supertrend(10,3) direction and line, keltner(20, 1.5×ATR20).
   • volatility: atr14, atr20.
   • structure: last confirmed fractal swing high/low (k=2) and when.
   • levels: priorRth high/low/close, overnight high/low, openingRange
     (first 15 min from 09:30 ET, only after it closes), vwapSession (18:00 ET
     anchor), vwapRth (09:30 ET anchor; null before today's 09:30 ET open
     and after 16:00 ET: use vwapSession then, as the strategy filters do).
   • vwap: which VWAP applies now (rth or session), the distance in
     ATR(14), and the RTH and session crosses in the last 30 bars.
   • participation: relative volume of the last bar and the last 3 vs the
     opening-range average, the 20 bars before, and the same time the
     previous trading day (orb's skip rule reads relVolLastVsOpeningRange).
   • Both count by trading day (from the 18:00 ET open): from 18:00 ET until
     the new range closes (09:30 ET + orbMinutes; 09:45 by default),
     levels.openingRange and the opening-range ratios are null, never the
     closed day's.
   • volumeProfile: computed from the bars, never estimate it.
     priorRth (the last complete RTH day, the day's reference), session
     (this Globex session, developing), rolling (the last vpLookback
     bars). Each: poc, vah, val, price (above, inside, or below
     value), fromPocAtr ((price - POC) / ATR(14)), the nearest high and low
     volume nodes (hvnAbove, hvnBelow, lvnAbove, lvnBelow), and all of
     them (hvn, lvn). With the default 100 rows levels fall between ticks:
     round to tickSize before using one as an order price. Bar-based: say so.
   • day (RTH only, else null): the open against the prior RTH day
     (openVs: inside the prior value area, outside value inside the range,
     or outside the range), the gap from the prior close, the openType
     (open-drive, open-test-drive, open-rejection-reverse, open-auction, from
     the first 30 minutes), the initial balance (ibHigh, ibLow, the first
     hour) and how far the day has gone past it (extUp, extDown), the
     dayType so far (inside the initial balance; normal variation: one side
     extended, under 2 IBs; trend: 2 IBs or more, or 1 from a narrow IB under
     0.35 ADR; neutral: both sides), and the range against the average daily range (adr, over
     adrDays, default 10). line is the same as the prompt's "day:" line.
   • liquidity: the last 4 swing highs and lows, equal highs/lows (within
     0.1 x ATR), and open fair value gaps.
   • context: the numbers the strategies' skip rules name: EMA 9/20
     crosses in the last 30 bars, bars of falling ADX, Keltner width vs its
     20-bar average (a squeeze is under 1), the band-close streak, SuperTrend
     flips in the last 20 bars, the last 5 bars' range in ATR, the session's
     high and low so far, and order flow (real: buy/sell volume recorded,
     else the bar-shape estimate; OFI over 1/3/5 bars, 5-bar delta, volume vs
     its 60-bar average). Read a skip rule's number here; never estimate it.
   • regime: primary (trend-up, trend-down, range, transition), volatility
     (high, normal, low), tags, and the metrics behind them (ADX, EMA(20) slope
     in ATRs, VWAP crosses in 30 bars, ATR vs its average).
   • referenceStop: distance and long/short stop prices. Round to tickSize.

5. Check the last bar's time (last.t, UTC): older than one bar of the timeframe
   (plus a minute) at the time you read it means stale data; say so and
   don't plan from it.

6. Never quote a number the snapshot did not produce or a tool did not return.

Examples

adx 24.1, adxSlope +3.2, regime trend-up
→ trending; strategy triggers come from `strategies_scan` (each STRATEGY.md's rules).

referenceStop.long = 21481.37 → round down to tick 0.25 → 21481.25

Timestamps are UTC; sessions and ranges are computed in America/New_York.
Fractal swings need strictly higher/lower neighbours, so equal highs are not
swings (treat them as liquidity; see liquidity-concepts).