skill-transaction-cost-calibration

name	skill-transaction-cost-calibration
description	Calibrate commissions, bid-ask spread, slippage, and simple market-impact assumptions from executions, quotes, or OHLCV data. Use when an agent needs an auditable transaction-cost model for backtest assumptions, execution review, or capacity analysis.
quantSkills	{"organization":"https://github.com/quantskills","repository":"quantskills/skill-transaction-cost-calibration","repository_url":"https://github.com/quantskills/skill-transaction-cost-calibration","project_type":"skill","collection":"execution-analysis","license":"GPL-3.0-only","category":"trader-research","tags":["transaction-cost","slippage","market-impact","execution"],"platforms":["claude-code","codex","cursor","hermes","openclaw"],"language":"zh-en","status":"stable","validation_level":"runnable","maintainer_type":"community","requires":[],"summary_zh":"从成交、盘口或 OHLCV 数据校准滑点、点差、手续费与参与率冲击成本。","summary_en":"Calibrate execution friction from fills, quotes, and bars with coverage and participation diagnostics."}

{
  "version": 1,
  "task": {
    "placeholder": "例如：校准这批成交的滑点和参与率冲击成本",
    "required": true
  },
  "fields": [
    {"key": "executions_csv", "type": "text", "label": "成交 CSV"},
    {"key": "quotes_csv", "type": "text", "label": "盘口 CSV（可选）"},
    {"key": "bars_csv", "type": "text", "label": "K 线 CSV（可选）"},
    {"key": "commission_bps", "type": "number", "label": "手续费 bps"}
  ],
  "prompt_template": "请处理任务：{{task}}；成交：{{executions_csv}}；盘口：{{quotes_csv}}；K线：{{bars_csv}}；手续费 bps：{{commission_bps}}。附件：{{#attachments}}"
}

Transaction Cost Calibration

Use this skill to estimate realized trading friction without treating one fixed bps value as universal. It uses only information available at or before execution, reports cost distributions, and fits a transparent participation-based impact curve.

Core Workflow

1. Read references/input_contract.md and identify whether quotes, bars, or only executions are available.
2. Parse timestamps with an explicit UTC assumption. Sort each symbol before as-of joining a quote or bar; never use a future reference.
3. Run scripts/calibrate_costs.py. Prefer quote mid and spread; use the latest prior bar close only as a documented fallback.
4. Review execution_costs.csv for signed slippage, spread, participation, and total cost. Review summary.json for coverage and outliers.
5. Use the fitted curve only inside the observed participation range. Report range, sample size, and commission convention with any backtest assumption.

Command

python scripts/calibrate_costs.py --executions fills.csv --quotes quotes.csv \
  --bars bars.csv --commission-bps 2.5 --output-dir cost_out

Use --demo --output-dir cost_out for a deterministic smoke test without network access.

Output Contract

• execution_costs.csv: one row per fill with as-of reference, signed slippage bps, spread bps, participation, commission, and total cost.
• cost_curve.csv: binned participation with median and p95 observed cost.
• summary.json: sample coverage, cost quantiles, impact coefficients, and warnings.

Keep signed slippage (price improvement is negative) separate from absolute cost. Do not extrapolate beyond observed participation without an explicit sensitivity assumption.

Boundaries

• Use this skill for execution friction calibration and backtest cost assumptions.
• Do not use it as a replacement for portfolio liquidation stress testing or generic backtest skills.
• Do not infer alpha, trade recommendations, or broker quality from costs alone.
• Follow references/source_boundary.md for permitted sources.