agentleFS
voyager67voyager67/JaneStreet-Agentic-Skills/skills/market-making/SKILL.md

market-making

Skill13 starsChanged 5 days agoApache-2.0View on GitHub

Design, analyze, or improve two-sided quoting and liquidity-provision strategies, including reservation prices, inventory skew, spread selection, adverse-selection filters, hedging, and quote lifecycle. Use for electronic market making and dealer-style quoting, not generic directional trading.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not equate maker rebates with positive expected value.

    line 31
  2. Distinctive:

    Avoid adverse-selection labels that leak post-fill prices into pre-trade features.

    line 32
  3. Distinctive:

    Do not infer queue position from L2 data as though exact FIFO rank were observed.

    line 33
  4. Distinctive:

    Never send live orders or change production limits as part of research without explicit authorization and the firm's approval process.

    line 35

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Market-Making Research and Quoting

Work as a market-making researcher. Treat quoting as a joint problem in expected spread capture, fill probability, adverse selection, inventory risk, hedge cost, and operational constraints. A theoretical quote is a starting hypothesis, not a production policy.

Workflow

  1. Specify the venue and product. Record tick/lot size, price-time or pro-rata priority, maker/taker fees and rebates, auction behavior, trading hours, position limits, hedge instrument, and whether the strategy is passive, aggressive, or mixed.
  2. Build the economics per fill. Decompose expected P&L into spread capture, fees/rebates, post-fill markout, hedge slippage, funding/borrow/carry, and inventory valuation. Keep realized P&L and forecast components separate.
  3. Establish a fair-value estimate. State the reference price and inputs. Compare mid, microprice, cross-venue fair value, and model value where relevant. Quantify stale-input and basis risk.
  4. Model inventory explicitly. Define signed inventory, target inventory, hard limits, risk horizon, volatility estimate, and penalty for carrying risk. A classical Avellaneda–Stoikov reservation-price approximation is r = S - q * gamma * sigma^2 * tau, with long inventory q > 0; it expresses a price skew, not a universal quote formula. The model's assumptions (diffusion price, arrival-intensity form, utility, horizon) must be tested before applying it.
  5. Set quote distances. Explain what determines each side's distance from fair value: spread target, volatility, fill intensity, inventory, queue priority, adverse-selection forecast, and fees. Do not quote from a single closed-form equation without examining its assumptions.
  6. Estimate fill and toxicity jointly. Segment fill probability and post-fill markout by quote distance, queue position, market state, side, and time horizon. A higher fill rate can be bad if it concentrates in informed flow.
  7. Define quote lifecycle. Specify new/replace/cancel decisions, minimum resting time if any, rate limits, stale-data behavior, cancel-on-disconnect, and what happens when acknowledgements lag.
  8. Evaluate inventory and hedging. Report inventory distribution, time at risk limits, tail exposure, hedge frequency, hedge cost, and P&L by inventory bucket. Include correlated instruments and basis risk where applicable.
  9. Run controlled experiments. Compare against a simple baseline, isolate one policy change at a time, and test across volatility/liquidity regimes and out-of-sample periods.

Required diagnostics

  • Bid and ask fill rates, fill-to-cancel ratio, queue-adjusted fill rate.
  • Effective spread, realized spread, signed markouts at multiple horizons.
  • Gross spread capture versus net economics after fees, rebates, hedge costs, and inventory P&L.
  • Inventory mean, dispersion, tails, limit breaches, and time-to-flat.
  • Quote age, cancel/replace rate, reject rate, acknowledgement delay, and stale-quote exposure.

Guardrails

  • Do not equate maker rebates with positive expected value.
  • Avoid adverse-selection labels that leak post-fill prices into pre-trade features.
  • Do not infer queue position from L2 data as though exact FIFO rank were observed.
  • Calibrate volatility and fill intensity on the same relevant instrument/session regime; disclose extrapolation.
  • Never send live orders or change production limits as part of research without explicit authorization and the firm's approval process.

Deliverable

Produce a quote-policy specification, assumptions table, unit-economics decomposition, fill/markout study, inventory analysis, regime robustness, operational failure behavior, and a clear go/no-go recommendation with evidence gaps.

Further reading

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.