@bitget-ai/getagent-skill 0.2.1

package/skills/getagent/references/package-schema.md

@@ -0,0 +1,552 @@
+# Playbook Package Schema
+
+## Contents
+
+- [Directory Layout](#directory-layout)
+- [manifest.yaml](#manifestyaml)
+- [backtest.yaml](#backtestyaml)
+- [src/main.py](#srcmainpy)
+
+## Directory Layout
+
+Implemented upload layout:
+
+```text
+<package-name>/
+├── README.md
+├── manifest.yaml
+├── src/
+│   ├── main.py
+│   ├── strategy.py
+│   ├── indicators.py
+│   ├── risk.py
+│   ├── execution.py
+│   ├── prompts.py        # optional
+│   ├── llm_policy.py     # optional
+│   └── agent_tools.py    # optional
+└── backtest.yaml         # optional; only for backtest_support: full
+```
+
+Design rules:
+
+- **Required:** `README.md`, `manifest.yaml`, `src/main.py`
+- **Optional in upload:** extra modules under `src/`, `backtest.yaml`
+- **Local-only (must not upload):** `tests/`, `notebooks/`, `research/`, `data/`,
+`backtest_results/`, `logs/`, `output/`, virtualenv and cache directories
+- **Do not infer strategy class from file layout.** Public behavior belongs in
+`manifest.yaml`.
+
+## README.md
+
+Human-readable strategy explanation. This file is required because subscribers
+need to understand the Playbook before trusting its signals.
+
+Write in plain language, not implementation notes. Avoid code blocks unless they
+make the explanation easier to understand. Cover at least:
+
+- What market behavior this strategy tries to capture
+- When it opens a long or short position
+- When it closes, takes profit, or stops loss
+- What each user-tunable parameter means and what happens when it is raised or
+  lowered
+- How to read the backtest metrics, especially strategy return vs account return
+- Main risks and market conditions where the strategy can perform poorly
+
+`README.md` must be UTF-8 text, at least 200 characters, and include enough
+plain-language content for the upload validator to detect sections covering
+策略、开仓、平仓、风险.
+
+## manifest.yaml
+
+Package identity, public behavior, and runtime contract.
+
+
+| Field                    | Type         | Required | Description                                                                                                                                   |
+| ------------------------ | ------------ | -------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
+| `name`                   | string       | yes      | URL-safe identifier. Lowercase, alphanumeric, hyphens only. 1-63 chars.                                                                       |
+| `display_name`           | string       | yes      | Human-readable name.                                                                                                                          |
+| `version`                | string       | no       | Server-assigned on publish. Do not write this for drafts; `upload` returns a `draft_id`, and `publish` returns the formal semver.              |
+| `description`            | string       | yes      | One-line description.                                                                                                                         |
+| `long_description`       | string       | yes      | 300–400 English words. Plain-language strategy summary that subscribers read in the marketplace UI. Must cover what the strategy captures, when it enters/exits, what each tunable parameter does, and where it underperforms — without revealing numeric parameters, indicator periods, thresholds, or formulas. See [`long_description` Writing Rules](#long_description-writing-rules). |
+| `market_type`            | string       | yes      | `"spot"` or `"contract"`.                                                                                                                     |
+| `trading_symbols`        | list[string] | yes      | Trading pairs, e.g. `["BTCUSDT"]`.                                                                                                            |
+| `decision_mode`          | string       | yes      | `deterministic`, `llm_assisted`, or `agentic`.                                                                                                |
+| `backtest_support`       | string       | yes      | `full` or `none`.                                                                                                                             |
+| `runtime_profile`        | string       | yes      | `deterministic`, `llm_bounded`, or `agentic`. Historical backtest uses `deterministic`; live-only bounded model execution uses `llm_bounded`. |
+| `execution_mode`         | string       | yes      | `signal_only` or `follow_trade`.                                                                                                              |
+| `follow_trade_supported` | bool         | yes      | Whether the Playbook can follow trades using a subscriber's bound subaccount.                                                                  |
+| `strategy_config`        | mapping      | no       | User-tunable strategy parameters read by both live and backtest code paths.                                                                    |
+| `user_config_schema`     | mapping      | no       | Authoritative declaration of which `strategy_config` fields users may override when subscribing.                                               |
+| `tags`                   | list[string] | no       | Searchable tags.                                                                                                                              |
+| `schedule.cron`          | string       | no       | Suggested publish-time schedule, e.g. `"0 */4 * * *"`.                                                                                        |
+| `schedule.tz`            | string       | required with `schedule.cron` | Default IANA timezone for subscription instances, e.g. `"Asia/Shanghai"`. Users may override it when creating their own instance. |
+| `official_evidence_kind` | string       | no       | For live-only strategies, optional declared evidence type: `paper` or `live`.                                                                 |
+
+
+**Name rules:** must match `^[a-z0-9]([a-z0-9-]{0,61}[a-z0-9])?$`.
+
+### Contract rules
+
+- Use `backtest_support: full` only when the core trading logic is fairly replayable on historical data.
+- Do not manually bump `version` in `manifest.yaml`. Versions belong to published
+  artifacts and are assigned by the server from the strategy's latest published
+  version.
+- Use `backtest_support: none` when open-ended LLM reasoning or unreplayable external context determines the trade.
+- `decision_mode: agentic` requires `runtime_profile: agentic`.
+- `runtime_profile: llm_bounded` requires `backtest_support: none`.
+- `execution_mode: follow_trade` requires `follow_trade_supported: true`.
+- Live-only playbooks must not declare `official_evidence_kind: backtest`.
+- Backtest-capable playbooks should not predeclare paper/live official evidence.
+- `schedule.cron` must not run more often than every 10 minutes. This limits
+  live execution frequency only; Playbooks may still use 5m or other finer-grain
+  bars for historical replay and signal features.
+- `schedule.tz` is required whenever `schedule.cron` is present. Use a valid
+  IANA timezone such as `Asia/Shanghai`. This is the default for new
+  subscription instances, not a strategy parameter, so do not put timezone in
+  `strategy_config` or `user_config_schema`.
+- Live contract order prices must align with exchange tick size. Do not pass
+  fixed decimal rounding such as `str(round(tp_price, 2))` to
+  `tp_trigger_price` / `sl_trigger_price`; use
+  `trade.helpers.resolve_contract_tpsl(...)` or quantize with
+  `trade.helpers.contract_rules(symbol).price_step`. The helper is
+  keyword-only and does not accept percentage override kwargs such as
+  `tp_pct_override` or `sl_pct_override`; convert percentage TP/SL tunables to
+  concrete trigger prices before calling it.
+- Live code should use `runtime.emit_signal_or_follow(...)` to emit the signal
+  and let the runtime decide whether to call the provided trade callback. In
+  `signal_only`, it returns successfully without placing orders.
+- Do not add `run_mode`. Historical vs live execution is selected by
+  `runtime.evaluation_mode`, injected by the platform at run time.
+- Every user-editable strategy parameter must appear in both `strategy_config`
+  (default runtime value) and `user_config_schema` (type/range/options for UI
+  and API validation). Do not declare fake editable fields that strategy code
+  does not read.
+- Playbook code must read editable values from
+  `runtime.manifest["strategy_config"]`; never hardcode user-editable values in
+  Python constants.
+- For configurable trading pairs, keep top-level `trading_symbols` and
+  `strategy_config.trading_symbols` aligned. The platform syncs the top-level
+  field when a user override changes `strategy_config.trading_symbols`.
+- **`strategy_config.margin_budget` is required** when `backtest_support: full`
+  or `execution_mode: follow_trade`. It is the per-strategy denominator the
+  platform uses to compute user-facing return % (`net_pnl / margin_budget`).
+  Validation rejects manifests that omit it or set it to a non-positive value.
+  Subscribers may override it via `config_overrides.margin_budget`; otherwise
+  the manifest default applies. Pure signal-only / live-only Playbooks that
+  never surface a return % may omit the field.
+
+Example:
+
+```yaml
+name: btc-ema-crossover
+display_name: "BTC EMA Crossover"
+description: "Trend following strategy based on EMA 12/26 crossover on 4h chart"
+long_description: |
+  This Playbook is a trend-following strategy on BTC perpetual futures. It is
+  built on the assumption that once the broader market enters a sustained
+  directional move, price tends to travel along that direction in a relatively
+  clean trend rather than chopping back and forth. The goal is to capture the
+  middle portion of those trends and stay out of grinding sideways markets where
+  most signals are noise.
+
+  Entry decisions are driven by alignment between a shorter-term and a
+  longer-term directional read. When both reads point clearly upward, the
+  strategy opens a long; when both point clearly downward, it opens a short. It
+  deliberately waits until momentum has formed and is still in motion, rather
+  than trying to fade extremes or call tops and bottoms.
+
+  Exits work by inversion. When the shorter-term direction flips before the
+  longer-term direction does, the strategy treats that as a fade in conviction,
+  retracts the trade signal, and closes the position. The intent is to lock in
+  whatever was captured rather than insist on every trade winning. Many small
+  losing trades are accepted as the cost of catching the occasional larger
+  trend run.
+
+  Two parameters are exposed to subscribers: leverage and margin budget. Higher
+  leverage amplifies both upside and drawdowns equally; the strategy does not
+  become more selective when leverage rises. Margin budget is the per-strategy
+  cap that the platform sizes orders against and uses as the denominator for
+  return percentage; treat it as the maximum amount of capital you are willing
+  to put at risk on this Playbook.
+
+  The strategy underperforms in choppy, range-bound markets where short-term
+  direction flips repeatedly without committing to a real trend. Gap-driven
+  moves around major events, illiquid pairs, and persistent funding-rate
+  dislocation can also produce a string of stops or trapped positions. Strong
+  historical backtest metrics are not a guarantee of live profitability — match
+  this Playbook against your own risk tolerance before subscribing.
+market_type: contract
+trading_symbols: ["BTCUSDT"]
+tags: ["trend", "ema", "btc", "contract"]
+schedule:
+  cron: "0 */4 * * *"
+  tz: "Asia/Shanghai"
+decision_mode: deterministic
+backtest_support: full
+runtime_profile: deterministic
+execution_mode: follow_trade
+follow_trade_supported: true
+
+strategy_config:
+  trading_symbols: ["BTCUSDT"]
+  fast_period: 12
+  slow_period: 26
+  leverage: 5
+  margin_budget: "50"
+
+user_config_schema:
+  trading_symbols:
+    type: array
+    item_type: string
+    default: ["BTCUSDT"]
+    options: ["BTCUSDT", "ETHUSDT", "SOLUSDT", "BNBUSDT", "XRPUSDT"]
+    aliases:
+      btc: BTCUSDT
+      bitcoin: BTCUSDT
+      eth: ETHUSDT
+      ethereum: ETHUSDT
+    min_items: 1
+    max_items: 1
+    label: "Trading symbol"
+  leverage:
+    type: integer
+    default: 5
+    min: 1
+    max: 20
+    aliases:
+      10x: 10
+      10倍: 10
+    label: "Leverage"
+  margin_budget:
+    type: string
+    default: "50"
+    pattern: "^[0-9]+(\\.[0-9]+)?$"
+    label: "Margin budget USDT"
+```
+
+Supported `user_config_schema.*.type` values are `string`, `integer`,
+`number`, `boolean`, and `array`. `array` may declare `item_type`, `min_items`,
+and `max_items`. `options` restricts scalar values or each array item.
+`aliases` maps common user phrases to canonical values. The platform also
+normalizes common raw user intent before validation: `eth` -> `ETHUSDT`,
+`10倍` / `10x` -> `10`, and `200U` / `200 USDT` -> `200` for budget fields.
+`key_aliases` or `synonyms` may be used when a field needs custom input names
+beyond built-ins such as `symbol` / `交易对` -> `trading_symbols` and
+`杠杆` / `倍数` -> `leverage`.
+
+### long_description Writing Rules
+
+`manifest.long_description` is the strategy summary that subscribers see in the
+marketplace UI when deciding whether to subscribe. Unlike `description` (one
+line) or `README.md` (full reference), it is a single short essay. Two goals
+have to coexist:
+
+1. **Make a non-technical user understand the strategy** — what market behavior
+   it bets on, when it enters and exits, what they can tune, and where it can
+   lose them money.
+2. **Protect the strategy** — never reveal the exact recipe a competitor or
+   reader could use to clone or front-run it.
+
+#### Length
+
+- Target **300–400 English words**.
+- Hard floor: **250 words**. Hard ceiling: **500 words**. Outside this range,
+  `validate.py` fails the upload.
+- Write in English only. Do not mix Chinese into the same field.
+
+#### Required 5-section skeleton
+
+The text must read as continuous prose (no bullet headers, no numbered lists),
+but must cover all five logical sections in order. Aim for 60–90 words per
+section.
+
+1. **What it does and what behavior it captures.** The strategy thesis. Which
+   market(s), which side (long/short/both), which kind of price behavior it
+   bets on.
+2. **How it decides to enter.** Direction-of-decision language only ("enters
+   long when momentum aligns with trend"). No periods, no thresholds, no
+   formulas.
+3. **How it decides to exit, take profit, or stop loss.** Same direction-of-
+   decision phrasing.
+4. **What subscribers can tune and what each tunable does.** Describe the
+   *effect* of raising or lowering each user-editable parameter declared in
+   `user_config_schema`. Do not show default values inside the prose.
+5. **Risks and unsuitable market conditions.** Be honest about where the
+   strategy underperforms or actively loses money. Past backtest performance
+   disclaimer should appear here.
+
+The validator confirms each section is present by detecting required keyword
+clusters (`captures`/`thesis`/`aim` for §1, `enter`/`open`/`long`/`short` for
+§2, `exit`/`close`/`stop`/`take profit` for §3, `parameter`/`tunable`/
+`leverage`/`margin`/`adjust` for §4, `risk`/`drawdown`/`underperform`/
+`unsuitable`/`lose` for §5). Missing any cluster fails validation.
+
+#### What you may write (allowed)
+
+- The strategy *category* in plain English: `trend-following`, `mean
+  reversion`, `momentum`, `breakout`, `range`, `arbitrage`, `grid`,
+  `event-driven`, `risk parity`, `pairs`.
+- The general *kind* of feature input, e.g. "price action", "volatility
+  regime", "open-interest behavior", "funding-rate behavior", "liquidation
+  pressure", "order-book imbalance".
+- The general *direction* of decisions ("waits for momentum to align with the
+  trend before entering", "fades extreme moves after volatility contracts").
+- The *effect* of each tunable parameter ("higher leverage amplifies both
+  upside and drawdowns equally", "raising the symbol universe spreads risk
+  across more pairs but increases data load").
+- The *qualitative* market regimes where it works/fails ("performs well in
+  sustained trends; underperforms in choppy ranges or around major news
+  gaps").
+
+#### What you must not write (forbidden — validator hard-fails)
+
+- **Numeric indicator periods.** `EMA 12`, `RSI 14`, `MA 200`, `ATR 1.5`,
+  `Bollinger 20`, `MACD 9`, `ADX 14`, `Stochastic 5`, `VWAP 30`, `MFI 14`,
+  `CCI 20`, `OBV 50`. Pattern: any indicator name immediately followed by a
+  number.
+- **Numeric lookback windows.** `14 bars`, `20 candles`, `5-period`, `30-day`,
+  `4-hour`, `15-minute`, `90 days`, `1 week`. Pattern: any number followed by
+  a time/candle/period unit.
+- **Explicit numeric thresholds.** `> 30`, `<= 0.7`, `>= 3%`, `drops below 5%`,
+  `volatility above 25%`. Any explicit comparison operator with a number, or
+  any standalone percentage with a number, is forbidden.
+- **Multipliers and ratios.** `1.5x ATR`, `2x volatility`, `3:1 reward/risk`.
+  Any `<number>x` or numeric ratio is forbidden.
+- **Formulas or pseudocode.** No `=`, no arithmetic expressions, no `if/then`
+  blocks, no parameter names from the source code.
+- **Default values for tunable parameters.** Do not mention specific defaults
+  even though they are public in `manifest.strategy_config`. Describe the
+  *effect* of changing the parameter, not its default.
+- **Specific provider/endpoint names.** No `data.crypto.futures.kline`, no
+  `Coinglass`, no `Binance funding_weighted`. Stay one level above the SDK.
+
+#### Example
+
+See the `btc-ema-crossover` manifest example above for a full 300–400 word
+`long_description` block that satisfies all rules.
+
+#### Common authoring mistakes
+
+- Repeating `description` verbatim or padding it to length. The validator
+  catches near-duplicate strings.
+- Writing in numbered list / bullet form ("**1.** Goal: …"). Keep it as
+  flowing prose.
+- Naming the indicators by acronym alone with no number (`uses EMA`, `uses
+  RSI`). The validator allows this — a clue is fine, an exact period is not —
+  but treat naked indicator-name name-dropping as a code smell. Prefer
+  category-level language ("uses moving-average crossovers", "uses momentum
+  oscillator").
+- Mentioning *backtest* configuration parameters (`max_backtest_symbols`,
+  `backtest_batch_size`). These are operational, not strategy semantics, and
+  do not belong in this field.
+- Any text that translates to "buy when X drops more than N percent". Rewrite
+  as "buys after a sharp washout move" without a number.
+
+## backtest.yaml
+
+Optional Nautilus replay specification. This file is only valid when
+`manifest.yaml` sets `backtest_support: full`.
+
+The runner exposes this file to Playbook code via `runtime.backtest_spec`.
+
+Sections:
+
+
+| Field                         | Type                     | Description                                                         |
+| ----------------------------- | ------------------------ | ------------------------------------------------------------------- |
+| `venue`                       | mapping                  | Venue name, account type, OMS type, and starting balances           |
+| `instrument` or `instruments` | mapping or list[mapping] | Explicit instrument definitions and `bar_type` declarations         |
+| `strategy`                    | mapping                  | Author strategy module/class entry plus config payload              |
+| `execution`                   | mapping                  | Optional replay window such as `start` / `end`                      |
+| `data_requirements`           | mapping                  | Optional replay data contract such as required enriched bar columns |
+
+
+Optional `data_requirements.required_bar_fields` lets a deterministic strategy
+declare extra normalized bar columns it expects in addition to core OHLCV.
+These names should use lower snake case to match the replay engine's normalized
+DataFrame columns, for example `quote_volume`, `trade_count`, or
+`taker_buy_volume`.
+
+Every backtest instrument must declare explicit `maker_fee` and `taker_fee`.
+Do not rely on zero-fee defaults. High-frequency Playbooks can look profitable
+or only slightly negative without fees, while real execution would pay a fee on
+every entry and exit. Use conservative exchange rates for the target venue, for
+example `maker_fee: "0.0002"` and `taker_fee: "0.0005"` for a simple Binance
+contract assumption.
+
+Data coverage is part of the package contract. If a strategy requires enriched
+columns such as `quote_volume`, `taker_buy_ratio`, open interest, funding, or
+liquidation data, author code must prove that the selected DataSDK
+endpoint/provider/symbol combination returns enough rows for the claimed replay
+window and the exact fields used by the strategy. Do not silently fill a missing
+decision feature with constants such as `0`, `0.5`, or `False` and still claim a
+valid backtest. If a feature is degraded, expose that in metrics/meta or fail
+the run with a clear error.
+
+Large multi-symbol backtests must batch data fetches instead of issuing one
+tight loop over every symbol and endpoint. If a Playbook scans many symbols
+across several historical datasets, expose controls such as
+`max_backtest_symbols`, `backtest_batch_size`, and
+`backtest_batch_sleep_seconds` in both `strategy_config` and
+`user_config_schema`. Default to a modest slice, for example 30 symbols in
+batches of 5 with a short sleep between batches, and report
+`symbols_requested`, `symbols_loaded`, batch settings, and skipped symbols in
+metrics/meta.
+
+When `backtest.yaml` declares more instruments than the data fetch actually
+loads, author code must filter the effective runtime spec to the loaded
+instrument ids before calling `backtest.run(...)`. The engine expects OHLCV data
+for every declared instrument and will fail if the spec still includes missing
+symbols. For full-universe evidence, split the symbol universe across multiple
+runs or a controlled workflow rather than making one default run fetch every
+symbol.
+
+Sizing is also part of the public contract. Backtest and live signal output
+should expose `notional_usdt`, `min_open_notional_usdt`, and `sizing_ok` when
+orders are sized from margin budgets. If the configured margin and leverage are
+too small for exchange lot size, tell the user to raise margin or leverage.
+
+When the strategy class defines `set_feature_frames(...)` or exposes a
+`feature_frames` attribute, the replay engine injects one full DataFrame per
+instrument ID with those extra columns preserved. Nautilus bar construction
+still consumes OHLCV, but the strategy can read richer columns from the
+injected feature frames.
+
+Use `getagent.backtest.prepare_frame()` / `getagent.backtest.build_feature_frame()`
+in author code to assemble those extra columns from multiple `getagent.data`
+datasets before calling `backtest.run(...)`.
+
+Minimal example:
+
+```yaml
+venue:
+  name: BINANCE
+  account_type: CASH
+  oms_type: NETTING
+  starting_balances:
+    - amount: 100000
+      currency: USDT
+
+strategy:
+  module: strategy
+  class: DemoStrategy
+  config_class: DemoStrategyConfig
+  config:
+    order_id_tag: "001"
+    trade_size: "0.01"
+
+instrument:
+  id: BTCUSDT.BINANCE
+  kind: spot
+  raw_symbol: BTCUSDT
+  base_currency: BTC
+  quote_currency: USDT
+  price_precision: 2
+  size_precision: 6
+  price_increment: "0.01"
+  size_increment: "0.000001"
+  lot_size: "0.000001"
+  maker_fee: "0.0002"
+  taker_fee: "0.0005"
+  bar_type: BTCUSDT.BINANCE-1-HOUR-LAST-EXTERNAL
+```
+
+Feature-aware example:
+
+```yaml
+venue:
+  name: BINANCE
+  account_type: CASH
+  oms_type: NETTING
+  starting_balances:
+    - amount: 100000
+      currency: USDT
+
+strategy:
+  module: strategy
+  class: PriceVolStrategy
+  config_class: PriceVolStrategyConfig
+
+instrument:
+  id: BTCUSDT.BINANCE
+  kind: spot
+  raw_symbol: BTCUSDT
+  base_currency: BTC
+  quote_currency: USDT
+  price_precision: 2
+  size_precision: 6
+  price_increment: "0.01"
+  size_increment: "0.000001"
+  lot_size: "0.000001"
+  maker_fee: "0.0002"
+  taker_fee: "0.0005"
+  bar_type: BTCUSDT.BINANCE-1-HOUR-LAST-EXTERNAL
+
+data_requirements:
+  required_bar_fields:
+    - quote_volume
+    - trade_count
+    - taker_buy_volume
+```
+
+## src/main.py
+
+Sandbox entry point. The runner executes the package as `python -m src.main`.
+`src/main.py` remains the required handoff into the runtime, but additional
+modules under `src/**` are first-class and encouraged.
+
+For packages with both live and backtest behavior, keep `manifest.yaml`
+immutable and route in code:
+
+```python
+from getagent import runtime
+from . import main_backtest, main_live
+
+if runtime.is_historical():
+    main_backtest.run()
+elif runtime.is_live():
+    main_live.run()
+else:
+    raise ValueError(f"unsupported evaluation_mode={runtime.evaluation_mode!r}")
+```
+
+Put shared data fetch and feature engineering in a neutral module such as
+`features.py`. `main_backtest.py` must not import live trading code, because
+historical runs should not require a configured trading identity.
+
+**Allowed imports:** `getagent`, `nautilus_trader`, `pandas`, `numpy`, `json`,
+`math`, `datetime`, `pathlib`, `asyncio`, `typing`, `dataclasses`,
+`collections`, `functools`, `re`, `decimal`, `statistics`, `itertools`, plus
+local modules under `src/`**.
+
+**Blocked imports:** `requests`, `httpx`, `trade_sdk`, `ccxt`, `os`, `subprocess`,
+`importlib`, `socket`, and other network/system/database libraries.
+
+**Output convention:** signals via `runtime.emit_signal()`, reports/artifacts via
+`/workspace/output/`, and public backtest charts via `backtest.generate_chart()`.
+
+## Nautilus strategy lifecycle
+
+The Playbook runner uses a NautilusTrader version where strategy lifecycle
+cleanup APIs are instrument-scoped. Do not call these methods without an
+instrument id:
+
+```python
+def on_stop(self) -> None:
+    if self.config.instrument_id is not None:
+        self.cancel_all_orders(self.config.instrument_id)
+        self.close_all_positions(self.config.instrument_id)
+```
+
+For multi-instrument strategies, iterate `instrument_ids`:
+
+```python
+def on_stop(self) -> None:
+    for instrument_id in self.config.instrument_ids:
+        self.cancel_all_orders(instrument_id)
+        self.close_all_positions(instrument_id)
+```
+
+Invalid patterns such as `self.cancel_all_orders()` or
+`self.close_all_positions()` are rejected by `scripts/validate.py` and
+the server-side upload validator because they fail at replay shutdown.