K-Line Trading Plan

Combine technical evidence and risk construction in one auditable workflow. This skill is read-only: never place, cancel, or amend an order unless the user separately invokes an execution skill and explicitly authorizes the action.

Choose the deliverable

• Analysis only: explain trend, structure, momentum, volatility, volume, derivatives, and invalidation. Do not invent account-specific sizing.
• Plan: add conditional entry, stop, targets, R:R, and scenario branches.
• Sized plan: also calculate units, notional, margin, and account risk. This requires account equity, maximum account risk, maximum margin allocation, and maximum leverage.

If the user requests a sized plan but omits risk inputs, collect market data first, then ask for all missing risk inputs in one message. Do not delay useful analysis while waiting for private account details. The user may provide a hypothetical account size instead.

1. Resolve intent and instrument

Infer what is safe to infer and state it. Resolve:

• instrument ID and product: spot, perpetual swap, or dated future;
• direction: long, short, neutral, or let evidence decide;
• holding horizon and decision timeframe;
• plan currency and contract type;
• risk inputs for sizing, if requested.

Use exchange-native instrument IDs. For OKX, examples are BTC-USDT, BTC-USDT-SWAP, and a listed dated-futures ID. Verify that the instrument exists before analysis. Options require contract-specific Greeks and payoff logic; provide analysis only unless those inputs are available.

2. Gather current evidence

Prefer read-only OKX market tools or the OKX CLI. Record source timestamps and never present cached or example values as current.

Minimum useful data:

1. ticker and 24-hour volume;
2. candles for the execution timeframe plus at least two higher timeframes;
3. RSI, MACD, moving averages, ATR, and volume context;
4. support/resistance and recent swing structure;
5. for derivatives: funding, open interest, and order-book spread/imbalance.

Optional macro, liquidation, long/short, or whale data may improve context but must be labeled by source. Missing optional data is not a zero signal. Exclude it, reduce coverage, and explain the limitation.

For timeframe selection, evidence normalization, and conflict handling, read references/analysis-framework.md.

3. Score evidence without hiding uncertainty

Organize observations into three pillars:

Assign each observed signal a score from -100 to +100 and a within-pillar weight based on relevance. Use scripts/signal_score.py to calculate the composite, coverage, and unavailable fields. Treat a low-coverage strong score as low confidence, not as a strong recommendation.

Do not double-count correlated evidence. For example, EMA alignment and MACD trend derived from the same candles should not each receive full independent weight.

4. Convert evidence into a conditional setup

A valid setup needs all of the following:

• Trigger: a measurable event or entry zone, not "buy now" by default.
• Invalidation: a price beyond market structure, adjusted for volatility.
• Targets: technically defensible levels with net R:R after costs.
• No-trade condition: the event that makes waiting preferable.

When the requested direction conflicts with evidence, show both the requested scenario and the evidence-led alternative. Do not force a directional plan. When higher and lower timeframes conflict, either reduce size/confidence or wait for alignment.

Read references/plan-framework.md when producing a plan or sized plan.

5. Size the plan deterministically

For linear spot and USDT-margined products, run scripts/risk_plan.py with the proposed levels and user limits. The script returns conservative, balanced, and aggressive tiers while enforcing:

• maximum loss at the stop, including estimated fees and slippage;
• maximum margin allocation;
• maximum leverage;
• optional maximum notional exposure.

Do not use the linear calculator for inverse contracts or options. For those, calculate from the exchange contract specification and clearly state the unit. Never infer private account equity or silently choose a loss limit.

6. Present an auditable result

Return, in this order:

1. Decision summary: bias, confidence, coverage, and whether the setup is active, conditional, or no-trade.
2. Evidence table: timeframe/source, reading, direction, and timestamp.
3. Setup: trigger/entry zone, invalidation, targets, and no-trade condition.
4. Three risk tiers: units, notional, margin, leverage, stop loss in account currency and percent, and net R:R to each target.
5. Scenario branches: bullish, bearish, and range/stand-aside conditions.
6. Data-quality note: unavailable or stale fields and how they affect confidence.
7. Risk notice: informational analysis, not investment advice; execution remains the user's decision.

Use exact values where data supports them and formulas/placeholders where it does not. Never fabricate a live price, signal, account balance, or fill.