AGENT SKILL
fleet-management
GPU fleet orchestration: worker registration, health monitoring, throughput tracking, and mode scheduling
- License 许可
- MIT
- 最近更新
- 2026-03-26
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- fleet-management
- version
- 1.0.0
- description
- GPU fleet orchestration: worker registration, health monitoring, throughput tracking, and mode scheduling
02 资源文件 14
- 参考文档Examples.md4.4 KB
- 参考文档METHODOLOGY.md16.2 KB
- 参考文档Protocol.md11.0 KB
- 参考文档references/mode-reference.md3.1 KB
- 参考文档Templates.md3.4 KB
- 参考文档Workflows/AnalyzeResults.md6.2 KB
- 参考文档Workflows/DefineGoal.md4.7 KB
- 参考文档Workflows/DesignExperiment.md8.3 KB
- 参考文档Workflows/FullCycle.md10.4 KB
- 参考文档Workflows/GenerateHypotheses.md5.9 KB
- 参考文档Workflows/Iterate.md7.6 KB
- 参考文档Workflows/MeasureResults.md6.4 KB
- 参考文档Workflows/QuickDiagnosis.md3.6 KB
- 参考文档Workflows/StructuredInvestigation.md6.0 KB
03 SKILL.md 正文
Fleet Management
Manage a distributed fleet of GPU workers (NVIDIA CUDA, Apple Metal, AMD HIP) performing Bitcoin private key search across 23 weak entropy modes.
Procedure
- Assess fleet status — Call
quantum_fleetto get active/offline workers, total throughput, and modes needing attention - Check health — Call
quantum_protectfor comprehensive health checks (worker connectivity, throughput, mode stalls) - Optimize priorities — If modes need rebalancing, call
quantum_optimizeto auto-adjust priority weights - Monitor costs — Call
quantum_costto identify inefficient modes consuming disproportionate GPU time - Review progress — Call
quantum_reviewfor overall fleet performance summary
Key Concepts
| Concept | Description |
|---|---|
| Worker | A GPU process running btc_puzzle with --worker flag, connected to the API |
| Mode | One of 23 weak key generation strategies (LCG, MT19937, SHA256-brain, etc.) |
| Priority Weight | 0.0-10.0 float controlling work distribution per mode (0 = disabled) |
| Work Unit | A keyspace range (mode + start_offset + end_offset) assigned to a worker |
| Checkpoint | A verifiable sample point within a work unit for anti-cheat validation |
| Canary | A honeypot target with a known private key, injected to detect cheating |
Mode Reference
| Mode | Name | Keyspace | Strategy |
|---|---|---|---|
| 0 | random | Infinite | Strong random (cuRAND) |
| 1 | time-lcg-glibc | ~63M | glibc LCG seeded by Unix timestamp |
| 2 | time-lcg-msvc | ~63M | MSVC LCG seeded by Unix timestamp |
| 3 | sha256-seq | 2^64 | SHA256(counter) |
| 4 | sha256-time | ~63M | SHA256(timestamp) |
| 5 | time-mt | ~63M | MT19937 seeded by timestamp |
| 6 | low-bits | 2^64 | key = counter (small keys) |
| 10 | openssl-pid | 65,536 | OpenSSL seeded by PID only |
| 12 | java-lcg | ~63B | Java LCG with ms timestamps |
| 13 | phone-sha256 | 10B | SHA256(phone_number) |
| 17 | v8-mwc1616 | ~4T | V8 JavaScript MWC1616 (Randstorm) |
| 20 | openssl-debian | 65,536 | Debian OpenSSL CVE-2008-0166 |
| 21 | sha256-short-ascii | ~744B | SHA256(short strings, len 1-8) |
References
04 同仓库更多 Skills
lucid-PLUGIN_NAME
PLUGIN_DESCRIPTION
★ 0
social-content
Generate short-form video content optimized for social media platforms
★ 0
personalized-outreach
Generate 1:1 personalized video for sales prospecting and customer success
★ 0
data-reports
Generate animated data report videos from KPIs, charts, and analytics
★ 0
lucid-video
Agent-native video generation using Remotion — MCP server + OpenClaw plugin
★ 0
lucid-veille
Content monitoring, AI digest generation, and multi-platform auto-publishing
★ 0