Skip to content

Solizardking/clawd-bot / .agents/skills/vulcan-scale-orders

vulcan-scale-orders

Unverified

Scale (laddered) limit orders: place a one-sided multi-limit entry across a price range, then attach laddered TP/SL. Suggests levels from technical analysis when the user does not provide a range.

vulcan-scale-orders

Use this skill for:

  • Entering a position with N limit orders distributed across a price range ("scale in" / laddered entry).
  • Exiting an existing position with laddered limits ("scale out").
  • Attaching laddered TP/SL to the resulting position after fills.
  • Generating sensible levels from TA when the user does not specify a range.

Do not use this skill for:

  • Two-sided market-making with automatic replacement on fill → use vulcan-grid-trading.
  • Time-sliced execution at the touch / mark price → use vulcan-twap-execution.
  • Single-shot limit or market orders → use vulcan-trade-execution.

Core Concept

A scale order places N post-only limit orders on one side between a lower_price and upper_price. Unlike a grid, there is no replacement loop: once a level fills, that slice of the entry is done. Profit and risk come from the eventual aggregate position, managed afterward with TP/SL.

Typical use cases:

  • Scale long entry: ladder buys from current price down through a support zone.
  • Scale short entry: ladder sells from current price up through a resistance zone.
  • Scale exit: ladder reduce-only sells above current (long) or buys below current (short).

Required Inputs

Collect with the user before placing:

FieldRequiredNotes
symbolyese.g. SOL
sideyesbuy for long entry / short exit, sell for short entry / long exit
lower_price, upper_priceyes (or TA-suggested)If omitted, derive from TA and ask the user to confirm before placing — see Prompting Pattern.
levelsyesNumber of orders (typical 3–8).
total_tokensyesTotal base-asset size across all levels. Converted to base lots per level.
distributionoptionaluniform (default), front_weighted, back_weighted, geometric:r=<ratio>. See Sizing Distributions.
tp_levels, sl_levelsoptionalLaddered exits to attach after fills. See Attaching TP/SL.
reduce_onlyoptionaltrue when scaling out of an existing position.
margin_modeoptionalcross (default for multi-limit) vs isolated. Isolated requires per-order placement — see Isolated Margin.

Before placing, choose a mode via vulcan-execution-modes (Paper / Dry-Run / Confirm-Each / Auto-Execute). Scale orders are typically a single batch placement, so Confirm-Each at the batch level (one approval covers the whole ladder) is standard.

Pick the mode before running any wallet-bound preflight. The Pre-Trade Checks below branch on mode — in paper or dry-run, never call vulcan_margin_status, vulcan_position_list, vulcan_trade_orders, or any other wallet-resolving tool. See Mode-Branched Preflight.

Prompting Pattern

The agent must prompt the user for missing or unconfirmed inputs before placing. This is non-negotiable for the price range — whether you suggested it from TA or the user gave one. Use the host's structured-question facility if it has one (e.g. AskUserQuestion); otherwise ask in plain text and wait for an explicit reply.

Decision tree

  1. User gave a complete range (lower_price + upper_price):

    • Restate it back in the confirmation table (Confirm With User).
    • Ask: "Confirm placing the ladder with lower=<L> / upper=<U> / levels=<N> / total_tokens=<T>? (yes / change / cancel)"
    • Do not silently accept; even a user-given range needs an explicit "yes" before placement (unless mode is auto-execute, where the plan is still echoed before placing).
  2. User gave a partial range (only one bound, only levels, only total_tokens, only "scale in around mark", etc.):

    • Fill the gaps from TA using the TA-Based Level Suggestion recipe.
    • Present the completed plan and ask the user to confirm or override each filled-in field.
  3. User gave no range at all ("scale in to SOL longs", "ladder a short on ETH"):

    • Run the TA suggestion recipe, present the suggestion compactly with the reasoning (which indicator drove each bound), and ask for confirmation.

Required confirmation question

After producing a suggested or restated plan, ask the user a single, focused question. Example phrasing:

Suggested scale-long range for SOL (1h TA):
  upper = 149.80   (mark 150.00 − 0.25·ATR)
  lower = 143.20   (BBand lower; ATR=2.7, recent swing low 142.8)
  levels = 5, spacing = 1.65, distribution = uniform, total = 2.5 SOL

  Sanity: ADX 18 (no strong trend), RSI 42 (neutral) — OK to scale long.

How would you like to proceed?
  1. Confirm as-is and place
  2. Tighten the range (e.g. only down to 145)
  3. Widen the range (e.g. down to BBand-2σ)
  4. Change levels / size / distribution
  5. Cancel

If a structured-question tool is available, use these as discrete options. If not, ask in chat and wait. Never place the ladder until the user answers option 1 (or supplies a modified plan that you then echo back for one more confirmation).

What "explicit approval" means

  • A bare "go", "yes", "do it", "place it", or "confirm" after you presented the full table counts.
  • "Sounds good, but tighten the lower bound to 145" is not approval — restate the modified plan and ask again.
  • Silence, an unrelated reply, or an emoji is not approval.
  • In auto-execute mode, the user's prior consent to auto-execute counts as approval for the placement, but you still echo the full plan immediately before placing and report the tx signature on completion.

Re-prompt triggers

Even after initial approval, re-prompt the user before placing if any of these are now true:

  • Mark price moved more than 0.5 × atr between suggestion and placement (the range may now be silly).
  • vulcan_margin_status.risk_state is not Healthy.
  • Existing position size on this symbol changed (new fills landed).
  • Any sanity warning newly fired (ADX crossed 30, RSI past extreme).

Show what changed and ask whether to proceed, adjust, or cancel.

Pre-Trade Checks

Run mode-agnostic market checks first — they are read-only and safe in every mode:

1. vulcan_market_info      → { symbol }    # base_lots_decimals, tick_size, min lot
2. vulcan_market_ticker    → { symbol }    # current price for centering / TA
3. vulcan_market_orderbook → { symbol }    # confirm side of book is empty above/below your range

Then ask the user for the execution mode (per vulcan-execution-modes). Only after the mode is set should you run wallet-bound checks.

Paper / Dry-Run: skip the wallet-bound block. The only paper-side check is:

vulcan_paper_status → {}     # if PAPER_NOT_INITIALIZED, propose `vulcan paper init --balance <N>`

Do not call vulcan_margin_status, vulcan_position_list, vulcan_trade_orders, vulcan_portfolio_*, or vulcan_wallet_list in paper/dry-run.

Confirm-Each / Auto-Execute: also run:

4. vulcan_strategy_preflight → { wallet }   # must report READY
5. vulcan_margin_status      → {}           # worst-case margin coverage
6. vulcan_position_list      → {}           # existing exposure in this market
7. vulcan_trade_orders       → { symbol }   # avoid colliding with existing resting orders

For buy ladders, upper_price must be at or below the best ask (or slide: true is required). For sell ladders, lower_price must be at or above the best bid. Otherwise post-only placement rejects.

TA-Based Level Suggestion (when user gives no range)

When the user says "scale in to SOL longs" with no prices, propose a range and confirm. Do not place until they approve.

Recipe:

  1. Get current mark from vulcan_market_ticker.
  2. Get a volatility / trend snapshot from vulcan_ta_report → { symbol, timeframe: "1h" }. Read off the per-indicator latest and signals blocks: bbands.latest.upper / .lower / .middle, atr.latest.atr (and atr.signals.atr_pct_of_price), rsi.latest.rsi (and rsi.signals.state), adx.latest.adx (and adx.signals.trend_strength). Don't request points_limit unless you actually need history.
  3. Pick the range:
    • Long entry (side: buy):
      • upper_price = current mark (or mark − 0.25 × atr for a small buffer below).
      • lower_price = max(bbands.latest.lower, mark − 2 × atr, recent_swing_low).
    • Short entry (side: sell):
      • lower_price = current mark (or mark + 0.25 × atr).
      • upper_price = min(bbands.latest.upper, mark + 2 × atr, recent_swing_high).
    • recent_swing_low / high: pull from vulcan_market_candles over the last 50–100 bars at the chosen timeframe.
  4. Pick levels from range width:
    • Spacing should be ≥ 0.25 × atr so fills aren't all clustered into noise.
    • levels = clamp( round( (upper − lower) / (0.5 × atr) ), 3, 8 ).
  5. Snap every price to the market tick_size from vulcan_market_info.
  6. Sanity gate before suggesting:
    • If adx.signals.trend_strength is strong or very_strong and the range is against the trend (e.g. scaling longs into a strong downtrend), flag this and ask the user to confirm — they may be catching a falling knife.
    • If rsi.signals.state is oversold (long entry) or overbought (short entry), the range may be too late / too early — surface it.

Present the suggestion compactly:

Suggested scale-long range for SOL (1h TA):
  upper = 149.80  (mark 150.00 − 0.25·ATR)
  lower = 143.20  (BBand lower; ATR=2.7, swing low 142.8)
  levels = 5, spacing = 1.65, distribution = uniform
  ADX 18 (no strong trend), RSI 42 (neutral). Confirm to place?

Calculating Levels

spacing = (upper_price − lower_price) / (levels − 1)

Buy ladder (descending — best price first is closest to current):

prices = [ upper, upper − spacing, upper − 2·spacing, …, lower ]

Sell ladder (ascending):

prices = [ lower, lower + spacing, …, upper ]

Every price must be a valid multiple of tick_size.

Sizing Distributions

tokens_per_level is then converted to base lots:

size_lots = tokens_per_level × 10^base_lots_decimals
  • uniform (default): tokens_per_level = total_tokens / levels at every price.
  • front_weighted: more size at the price closest to current mark — conservative; better fill probability, worse average if the range fully fills.
  • back_weighted: more size at the far end (lower price for buys, higher for sells) — aggressive; pays off if price runs deep into your range, but tail risk if not all fill.
  • geometric:r=R (R != 1): each level's size is R × the previous level's. R = 1.5 for back-weighted (long entry into deep support), R = 1/1.5 for front-weighted. Pick R so the sum matches total_tokens:
    s0 = total_tokens × (1 − R) / (1 − R^levels)   # geometric series scaling
    size_i = s0 × R^i                              # i = 0 .. levels−1
    

If any computed size_lots < min_lot_size from vulcan_market_info, reduce levels or increase total_tokens and re-suggest. Never silently round to zero.

Margin Estimation

Worst case is all levels fill → full total_tokens long or short at the average ladder price.

avg_fill_price ≈ Σ (price_i × size_i) / Σ size_i
notional      ≈ avg_fill_price × total_tokens
initial_margin = notional / max_leverage_at_this_size

Get max_leverage_at_this_size from vulcan_margin_leverage_tiers for the resulting position size. Confirm vulcan_margin_status.available_collateral ≥ initial_margin × safety_buffer (1.2×) before placing.

Route any leverage / margin warnings through vulcan-risk-management.

Confirm With User

Always present the full plan as a compact table before placing:

Scale long SOL — 5 levels, uniform, total 2.5 SOL

  level  price    tokens   lots    cum_tokens
  1      149.80   0.50     50      0.50
  2      148.15   0.50     50      1.00
  3      146.50   0.50     50      1.50
  4      144.85   0.50     50      2.00
  5      143.20   0.50     50      2.50

  Worst-case avg fill ≈ 146.50, notional ≈ $366.25
  Worst-case initial margin ≈ $73 (5× tier), available collateral $480 ✓
  No TP/SL attached at order time — laddered TP/SL will be set post-fill.

Wait for explicit approval. In auto-execute mode, log the same plan immediately, then place, and report the tx signature as soon as it lands.

Place The Ladder

Single batch placement, post-only:

vulcan_trade_multi_limit → {
  symbol: "SOL",
  bids: [
    { price: 149.80, size: 50 },
    { price: 148.15, size: 50 },
    { price: 146.50, size: 50 },
    { price: 144.85, size: 50 },
    { price: 143.20, size: 50 }
  ],
  asks: [],
  slide: false,
  acknowledged: true
}
  • For a sell ladder, put the orders in asks and leave bids empty.
  • slide: false enforces strict post-only at the requested price — preferred for scale entries. Use slide: true only when the user explicitly accepts repricing to the top of book on collision.
  • vulcan_trade_multi_limit accepts optional per-leg tp and sl on each bid/ask entry. When any leg carries tp/sl the submission expands to N individual limit-order instructions bundled into a single transaction (each with its own bracket); slide is ignored on that path. See Attaching TP/SL.

Verify placement immediately:

vulcan_trade_orders → { symbol: "SOL" }

Report every accepted order plus the full transaction signature.

Attaching TP/SL

vulcan_trade_multi_limit supports native per-leg TP/SL via optional tp and sl on each bid/ask entry. When any leg carries tp/sl the call still submits in one transaction (Phoenix's multi_limit instruction cannot bundle conditionals itself, so the SDK expands the batch to N individual place_limit_order_with_conditionals ixs in the same tx). slide is ignored on that path; entries are post-only.

Three TP/SL models exist for a scale ladder:

  • Per-leg brackets (default, recommended) — each ladder leg ships with its own tp / sl in the same vulcan_trade_multi_limit tx. One signature, one tx, each leg's bracket sized to that leg's base lots and armed when that leg fills.
  • Position staircase — place the ladder with no per-leg tp/sl, then attach tp_levels / sl_levels on the resulting position via vulcan_trade_set_tpsl. The exit fires off mark vs. the aggregated position size, so this is the right choice when you want one unified staircase across the new ladder and any pre-existing position in the same symbol.
  • Deferred — no TP/SL during entry. Only acceptable when the far end of the ladder is itself the hard invalidation. Risky and discouraged.

Choosing the model

Before placing, if TP/SL is in scope, pick using this rule and confirm with the user:

  • No existing position in this symbol → per-leg brackets.
  • Adding to an existing position and the user wants one unified exit ladder → position staircase.
  • User explicitly accepts running without an SL during entry and the bottom of the ladder is their invalidation → deferred.

If you ask the user, present the three options in that order with one-line descriptions; do not re-introduce the A/B/C labels.

Per-leg brackets (default)

Pass tp and sl on each bid/ask entry. One transaction, one signature, each leg's bracket sized to that leg's base lots:

vulcan_trade_multi_limit → {
  symbol: "SOL",
  bids: [
    { price: 149.80, size: 50, tp: 157.29, sl: 145.31 },
    { price: 148.15, size: 50, tp: 155.56, sl: 143.71 },
    { price: 146.50, size: 50, tp: 153.83, sl: 142.11 },
    { price: 144.85, size: 50, tp: 152.09, sl: 140.50 },
    { price: 143.20, size: 50, tp: 150.36, sl: 138.90 }
  ],
  asks: [],
  acknowledged: true
}

Per-leg prices typically use a fixed offset from each leg's entry, e.g.:

tp_for_this_leg = price × (1 + tp_pct)         # long ladder
sl_for_this_leg = price × (1 − sl_pct)

Constraints and gotchas:

  • Cross margin only. Multi_limit (with or without brackets) rejects isolated-only markets. For isolated-margin scale entries, fall back to per-order vulcan_trade placement and accept the N-transaction cost.
  • slide: true is ignored when any leg has tp/sl. The expanded per-leg limit ixs use Phoenix's standard post-only placement; explicit slide isn't exposed by the limit-order builder.
  • No explicit-size brackets on limits. The SDK rejects BracketLegSize::BaseLots on limit-leg brackets (UnsupportedLimitBracketLegSizing); the on-chain max_size defaults to the leg's base lots, which is what you want here.
  • TP/SL roll-back rule still applies. If a bracket would reduce (rather than open or extend) the position when it fires, that trigger rolls back. Per-leg brackets are safe for the entry phase but never for scale-out — use the position staircase tp_levels instead.
  • No partial-fill brackets. If a leg only partially fills, the bracket still arms for the full leg size, which can exceed what actually filled. Re-fetch vulcan_position_show after the tx lands and reconcile.
  • Brackets only cover the new ladder. If a prior position in the same symbol already exists, per-leg brackets do not cover it. Use the position staircase instead when you want one unified exit.

Confirm direction rules from vulcan-tpsl-management for every computed pair before placing.

Report the tx signature and the final per-leg table as soon as the multi-limit lands; surface each fill (and any bracket trigger) immediately — never batch the updates.

Position staircase (when adding to an existing position)

Place the ladder with multi_limit (no per-leg tp/sl), wait until at least one level fills (position exists), then attach laddered TP/SL on the position:

vulcan_trade_set_tpsl → {
  symbol: "SOL",
  tp_levels: [
    { price: 158.0, size: 1.0 },   # close 1.0 SOL at $158
    { price: 165.0, size: 1.0 },   # close 1.0 SOL at $165
    { price: 175.0, size: 0.5 }    # runner — last 0.5 SOL at $175
  ],
  sl_levels: [{ price: 138.0 }],   # full position at $138
  acknowledged: true
}

Constraints (from vulcan-tpsl-management):

  • Sum of tp_levels sizes must not exceed current filled position size. As more entry levels fill, you may need to append additional TP slices with another vulcan_trade_set_tpsl call (set_tpsl appends; it does not replace).
  • Long ladder → TP above entry, SL below. Short ladder → TP below entry, SL above.

Direction sanity check

Before calling vulcan_trade_set_tpsl, re-fetch position side and entry price:

vulcan_position_show → { symbol: "SOL" }

This avoids attaching wrong-direction TP/SL if the user changed plans between ladder placement and fill.

Deferred TP/SL (discouraged)

Simpler but riskier — the position is unprotected while levels fill. Only acceptable if:

  • The user explicitly accepts running without SL during the entry, and
  • The full lower_price (or upper_price) is itself a hard invalidation level the user is willing to ride down to.

If the user picks deferred, set a wakeup or schedule a follow-up check; do not just walk away.

Scale-Out (reduce-only)

To scale out of an existing long, use a sell ladder with reduce_only: true. vulcan_trade_multi_limit does not accept per-order reduce_only. Two options:

  1. Prefer: use multi-level TP via tp_levels on the existing position (see Position staircase) — these are reduce-only trigger orders by construction and survive position-side changes safely.
  2. Fallback: if the user wants resting limit sells (not triggers) above market for a long, place them individually:
    for price in ladder:
      vulcan_trade → {
        symbol, side: "sell", order_type: "limit",
        price, size: <lots>, reduce_only: true, acknowledged: true
      }
    
    Slower than multi_limit but the only way to get reduce-only resting limits today.

Default to option 1 unless the user explicitly asks for resting reduce-only limits.

Isolated Margin

vulcan_trade_multi_limit submits a single cross-margin transaction. For isolated-margin scale entries, fall back to per-order placement:

for (price, lots) in ladder:
  vulcan_trade → {
    symbol, side, order_type: "limit",
    price, size: lots,
    isolated: true,
    collateral: <only on the first order>,
    acknowledged: true
  }

Report each placed order and signature as it lands — do not batch the updates.

Monitoring & Cleanup

Scale orders have no replacement loop. After placement:

  1. Watch fills with vulcan_position_show → { symbol } and vulcan_trade_orders → { symbol }.
  2. If you chose the position staircase, top up tp_levels as fills come in so unprotected size never grows. (Per-leg brackets handle this automatically — each bracket arms with its own leg.)
  3. If the user changes their mind mid-fill, cancel remaining levels with:
    vulcan_trade_cancel → { symbol, scope: "all", acknowledged: true }
    
    then either close the partial position (vulcan_position_close) or attach the agreed TP/SL.
  4. If the range fully fills, confirm the final laddered TP/SL covers the entire position size — sum mismatches between filled size and tp_levels total are a common bug.

Report every fill, cancellation, and signature immediately — never batch into an end-of-run summary.

Hard Rules

  1. Never place a scale ladder without explicit user approval of the full level table (unless in auto-execute mode, which still requires immediate reporting).
  2. Always validate against tick_size and minimum lot size from vulcan_market_info — never silently round to zero on small far-end slices.
  3. Always compute worst-case (all levels fill) margin before placing.
  4. When TA-suggesting a range, surface trend / overbought-oversold warnings (adx > 30 against trend, rsi past extreme) and ask the user to confirm.
  5. Buy ladders sit at or below the best ask; sell ladders at or above the best bid — otherwise use slide: true only with explicit approval.
  6. Default to per-leg brackets. vulcan_trade_multi_limit supports native per-leg TP/SL via optional tp / sl on each entry — one tx, one signature, each leg's bracket sized to that leg's base lots. Use the position staircase (vulcan_trade_set_tpsl with tp_levels / sl_levels) only when adding to an existing position you want managed under one unified exit ladder. Re-fetch position side with vulcan_position_show before any set_tpsl call. Never claim TP/SL is active until it is recorded on the position.
  7. Sum of tp_levels sizes must never exceed currently filled position size; top up incrementally as more entry levels fill.
  8. Route errors by .error.category via vulcan-error-recovery. On tx_failed, re-fetch vulcan_trade_orders before retrying — partial placement is possible.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 192kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 192kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 192kUpdated Claude CodeCodexVerifiedInstall