跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
金融

create-strategy

Guide for creating a new Etemaro LP strategy from scratch or adapting an existing preset. Use this skill whenever a user asks to "create a strategy", "add a new strategy", "set up a strategy", "configure LP strategy", "what strategy should I use", "tweak my strategy", "change my strategy", or any variant involving Etemaro strategy setup, customization, or deployment configuration. Also trigger when the user describes a trading goal (e.g. "I want to go bullish on SOL", "I want to earn fees on a stable pair", "I have a volatile token") and needs it translated into a concrete config. Strategies are immutable by default: this skill always creates a NEW strategy (new id) and a NEW config file, and never edits or overwrites an existing strategy without explicit user confirmation.

etemaro · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add romankurnovskii/etemaro --skill create-strategy

技能说明


name: create-strategy description: > Guide for creating a new Etemaro LP strategy from scratch or adapting an existing preset. Use this skill whenever a user asks to "create a strategy", "add a new strategy", "set up a strategy", "configure LP strategy", "what strategy should I use", "tweak my strategy", "change my strategy", or any variant involving Etemaro strategy setup, customization, or deployment configuration. Also trigger when the user describes a trading goal (e.g. "I want to go bullish on SOL", "I want to earn fees on a stable pair", "I have a volatile token") and needs it translated into a concrete config. Strategies are immutable by default: this skill always creates a NEW strategy (new id) and a NEW config file, and never edits or overwrites an existing strategy without explicit user confirmation. metadata: version: 2.0.0 author: etemaro license: MIT

Create Strategy — Etemaro

This skill walks an agent through creating or selecting an LP strategy for the Etemaro trading daemon, then generates a ready-to-use config patch and (when needed) a new strategy-library entry.


Strategy Immutability (Non-Negotiable)

Strategies and their config files are immutable by default.

  • "Create a strategy" ⇒ always create a NEW strategy. New unique library id, new versioned config file. Never rename, overwrite, or edit an existing entry or config.
  • Applies to: entries in config/shared/strategy-library.json, and config files under config/instances/.
  • Bundled presets are read-only. config/shared/strategy-library.shared.json (and its source of truth packages/core/src/domain/strategy-library-shared.ts) ship with the repo. Never edit them in place — copy to a new private entry instead.
  • If the agent believes an existing strategy needs to change (market conditions shifted, performance is bad, a filter is wrong):
    1. STOP. Do not edit.
    2. State the exact field(s), current value(s), and proposed value(s).
    3. Give the reason and evidence (logs, metrics, file:line).
    4. Ask for explicit user confirmation.
    5. Recommend creating a new variant (copy the old entry/config → new id + new config file) so the original stays intact and comparable. This is the default suggestion, even with confirmation.

Only exception: missing / malformed required data

When a strategy or config is missing required data or is malformed, fix the minimum needed to restore validity — no separate confirmation gate, but report exactly what changed:

  • Missing required string id or name.
  • Invalid JSON / a parse error.
  • strategy.activeStrategyId pointing at a non-existent library id.
  • Active strategy missing smartWalletListId while the smart-wallet entry source is active.
  • Legacy snake_case keys (lp_strategy, token_criteria, exit.take_profit_pct, …).
  • Missing screening.market.enabled / screening.smartWallets.enabled so no entry source resolves.

While fixing, do not change intent or behavior — only restore the missing/broken data.

Decision flow

User asks to change an existing strategy/config?
  ├─ missing or invalid required data?  → fix in place (minimal), report the change
  └─ otherwise                          → propose change → ask confirmation → recommend NEW variant

User asks to create or tune a strategy?
  → always create a NEW library id + a NEW config file

Why this matters

Library entries are mostly metadata — see the next section. Editing an entry in place usually does not change bot behavior; behavior lives in the agent config. So "changing" a strategy means shipping a new config + a new entry, which also preserves the ability to compare runs.


What Actually Controls Runtime Behavior

Read this before "adjusting a strategy" — it determines where a change has to go.

SettingWhereEffect
strategy.activeStrategyIdagent configSelects the active library entry (context + smart-wallet list). USED
strategy.strategyMeteoraagent configActual deploy shape: spot | curve | bid_ask. USED
strategy.minBinsBelow / maxBinsBelow / defaultBinsBelow / minSafeBinsBelowagent configBin range below active bin. USED
management.*agent configStop-loss, take-profit, trailing, sizing, cooldowns. USED
screening.common / market / smartWalletsagent configPool filters and entry source. USED
opportunity.*agent configOpportunity poller + smart-wallet score bonus. USED
risk.maxPositions / maxDeployAmountagent configGlobal capital caps. USED
smartWalletListIdlibrary entryStartup validation + smart-wallet screening. USED
lpStrategy, tokenCriteria, range, exit.takeProfitPct, bestFor, rawlibrary entryStored / descriptive. condition, notes, bestFor feed the LLM strategy context; the rest are not read at deploy.

There is no bins_above config field and no ratio field: directional bias is expressed through strategyMeteora (shape) plus the *BinsBelow counts.


Phase 1: Gather Minimum Required Information

Collect the following from the user. Ask as a single, concise message — don't split into separate turns unless an answer is ambiguous. Extract anything already given in the conversation and ask only for what is missing.

Checklist (all required)

  1. Directional view — bullish, bearish, or neutral on the token?
  2. Token type — volatile/narrative, stable/high-volume pool, or blue-chip (SOL, ETH)?
  3. Risk tolerance — conservative (tight stop-loss, quick exits), moderate, aggressive (hold through dips, re-seed)?
  4. Capital per position (SOL) — how much SOL per position? (Drives deployAmountSol / positionSizePct.)
  5. Exit preference — take profit at a %/threshold, trail the price, or let it run and re-seed?
  6. Smart-wallet tracking — gate en
阅读完整来源 ↗