← 返回 Skills 市场
captainslab

Btc Sprint Stack

作者 Captain · GitHub ↗ · v0.4.2 · MIT-0
cross-platform ⚠ suspicious
140
总下载
0
收藏
0
当前安装
6
版本数
在 OpenClaw 中安装
/install btc-sprint-stack
功能描述
Trade BTC 5m/15m Polymarket fast markets on Simmer with dry-run-first execution, fee-aware filtering, bankroll limits, flat signal_data, journaling, heartbea...
使用说明 (SKILL.md)

BTC Sprint Stack

Use this skill to operate a conservative BTC 5m/15m sprint bot on Simmer.

This is a template. The default signal is short-horizon BTC momentum plus Simmer context filters. Replace the signal source, confidence model, or edge inputs without changing the execution, journaling, and risk plumbing.

Operating constraints

  • Default to dry-run.
  • Use the official SimmerClient from simmer-sdk.
  • Only target BTC fast markets (5m, 15m) from Polymarket.
  • Enforce risk defaults from config/defaults.json.
  • Merge learned tunables from data/live_params.json before env overrides.
  • Every trade must include:
    • source
    • skill_slug
    • reasoning
    • signal_data.edge
    • signal_data.confidence
    • signal_data.signal_source
  • LLM decisions must validate to strict JSON and remain BTC-only.

Entrypoint

./.venv/bin/python skills/btc-sprint-stack/main.py --once --dry-run --validate-real-path

Files

  • main.py — orchestration
  • modules/btc_sprint_signal.py — momentum and fallback signal
  • modules/btc_regime_filter.py — time, spread, edge, confidence, fee checks
  • modules/btc_sprint_executor.py — dry-run/live execution wrapper
  • modules/btc_position_manager.py — bankroll and position sizing
  • modules/btc_trade_journal.py — JSONL journal
  • modules/btc_self_learn.py — bounded parameter suggestions
  • modules/btc_heartbeat.py — run summary and briefing
  • modules/btc_llm_decider.py — strict JSON decision layer, provider abstraction, and learning store helpers
  • data/live_params.json — learned tunables overlay
  • data/pending_rules.json — pending and applied rule suggestions
  • data/llm_decisions.jsonl — per-cycle LLM decision log
安全使用建议
What to check before installing/running: 1) Expect to provide a SIMMER_API_KEY and some LLM credentials (or select an LLM provider). The package's internal manifest (clawhub.json) requires these even though the top-level registry metadata omitted them — treat that as a red flag. 2) The code will make network calls (Simmer, Binance, LLM providers, Discord if you enable webhooks) and will read/write files in the skill directory (data/*.jsonl). 3) The LLM layer may read additional credential locations: ~/.codex/auth.json or CODEX_OAUTH_TOKEN, and/or use Google ADC (google-auth) which can access host Google credentials; the skill also documents performing wallet linkage if WALLET_PRIVATE_KEY is set. Do not set WALLET_PRIVATE_KEY or point the skill at your primary cloud or codex credentials unless you understand the consequences. 4) Run only in dry-run mode first and in an isolated environment (dedicated test API keys, limited permissions, or a disposable VM/container). 5) If you plan to use live trading: create least-privilege API keys for Simmer, avoid using personal Google or Codex host credentials, and review data files and logs regularly. 6) The mismatch between declared required env vars in the registry vs. clawhub.json suggests the package metadata may be incomplete or out of date — ask the author to clarify required env vars and credential flows before enabling live operation.
功能分析
Type: OpenClaw Skill Name: btc-sprint-stack Version: 0.4.2 The BTC Sprint Stack is a trading bot bundle that exhibits high-risk capabilities, including the handling of 'WALLET_PRIVATE_KEY' for on-chain transactions and the retrieval of OAuth tokens from the user's home directory ('~/.codex/auth.json' in 'modules/btc_llm_decider.py'). It performs extensive network activity, connecting to Binance for price data, Discord for alerts, and multiple LLM providers for decision-making. While these behaviors are aligned with the stated purpose of an automated trading bot, the combination of private key handling, access to sensitive local configuration files, and the use of non-existent model identifiers (e.g., 'gpt-5-mini', 'gemini-2.5-pro') warrants a suspicious classification under the provided guidelines.
能力标签
cryptorequires-walletrequires-oauth-token
能力评估
Purpose & Capability
The code implements a BTC-only Simmer trading bot (signals, regime filtering, execution, journaling, LLM decision layer) which matches the skill name and description. The package's internal clawhub.json declares pip deps (simmer-sdk, eth-account, google-auth, requests) and env vars (SIMMER_API_KEY, LLM_API_KEY), which are reasonable for a trading+LLM integration. However the registry metadata/skill summary shown earlier reported no required env vars or credentials — an inconsistency between declared registry metadata and the packaged manifest.
Instruction Scope
SKILL.md and main.py expect to call external services (Simmer API, Binance public API, LLM providers) and to read/write local data files (data/*.jsonl). The runtime will insist SIMMER_API_KEY is present (main.py will exit otherwise). The LLM layer supports multiple provider flows and may read host auth files or environment tokens (see below). The SKILL.md does not enumerate all env vars and local credential paths the code will use, granting the skill broader scope than documented.
Install Mechanism
There is no explicit install spec in the top-level manifest delivered to the platform, but clawhub.json lists pip dependencies (simmer-sdk, eth-account, google-auth, requests). Those packages are plausible for the bot. The lack of a single explicit install section in SKILL.md/manifest is an inconsistency to be aware of (platform may attempt to install pip deps based on clawhub.json).
Credentials
The code requires SIMMER_API_KEY and some LLM credentials (clawhub.json lists LLM_API_KEY). In addition the LLM provider implementation will accept/attempt multiple credential sources: CODEX_OAUTH_TOKEN or ~/.codex/auth.json (reads user's home), Google Application Default Credentials via google-auth (which reads host ADC files), and optional WALLET_PRIVATE_KEY (the code documents wallet linkage/USDC.e approval when present). Those additional credential paths are not prominently documented in SKILL.md but allow access to host-stored secrets or private keys — this is disproportionate unless you intentionally enable those providers/keys.
Persistence & Privilege
The skill is not force-enabled (always:false) and does not request system-wide configuration changes. It will create and update local files under its data/ directory (journal.jsonl, llm_decisions.jsonl, live_params.json, pending_rules.json). The potential to read ~/.codex/auth.json and to perform wallet linkage when WALLET_PRIVATE_KEY is set is notable but the skill does not modify other skills or global agent settings.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install btc-sprint-stack
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /btc-sprint-stack 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v0.4.2
Fix: switch to api.binance.us — api.binance.com returns 451 geo-block from EU servers
v0.4.1
Fix: call auto_redeem() in live mode for all wallet types (not just external)
v0.4.0
Switch to OpenRouter free tier (openrouter/free auto-routes to best available free model)
v0.3.0
Add OpenAI Codex OAuth LLM provider (reads ~/.codex/auth.json)
v0.2.0
Polygon external wallet support + Google Cloud OAuth LLM provider
v0.1.0
Initial release
元数据
Slug btc-sprint-stack
版本 0.4.2
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 6
常见问题

Btc Sprint Stack 是什么?

Trade BTC 5m/15m Polymarket fast markets on Simmer with dry-run-first execution, fee-aware filtering, bankroll limits, flat signal_data, journaling, heartbea... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 140 次。

如何安装 Btc Sprint Stack?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install btc-sprint-stack」即可一键安装,无需额外配置。

Btc Sprint Stack 是免费的吗?

是的,Btc Sprint Stack 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Btc Sprint Stack 支持哪些平台?

Btc Sprint Stack 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Btc Sprint Stack?

由 Captain(@captainslab)开发并维护,当前版本 v0.4.2。

💬 留言讨论