技能 when-training-neural-networks-use-flow-nexus-neural
📦
when-training-neural-networks-use-flow-nexus-neural
內容修訂 r1 中風險 🔑 環境變數⚙️ 外部命令🌐 網路存取
使用 Flow Nexus 訓練神經網路
分散式神經網路訓練需要協調一致的設定、驗證與部署。此技能會引導 Flow Nexus 工作流程，從架構設計到生產環境服務。
支援: Claude Codex Code(CC)
使用我的智慧代理安裝
將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。
智能體請求
Review the Skillstore skill "when-training-neural-networks-use-flow-nexus-neural" from https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural.md and its manifest at https://skillstore.io/api/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。
Agent 可讀資源
當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。
測試它
正在使用「when-training-neural-networks-use-flow-nexus-neural」。 使用 Flow Nexus 訓練 transformer 分類器，並準備將其用於服務。
預期結果:
一份分階段計畫，涵蓋驗證、叢集設定、架構選擇、分散式訓練、驗證指標與部署就緒狀態。
正在使用「when-training-neural-networks-use-flow-nexus-neural」。 審查我的已訓練模型是否已準備好投入生產環境。
預期結果:
一份驗證摘要，包含準確率、延遲、過度擬合檢查、基準測試狀態與建議的部署動作。
正在使用「when-training-neural-networks-use-flow-nexus-neural」。 為已完成的神經網路專案建立部署交接資料。
預期結果:
一份交接大綱，包含模型中繼資料、服務端點、監控預期、健康檢查與營運注意事項。
安全審計
中風險v8 • 2026/7/5 開啟版本化報告
Most static detections are Markdown fences, workflow diagrams, local URLs, or illustrative labels rather than malicious code. Confirmed concerns are documented commands that may execute unpinned npx packages and a deployment example that can expose an unauthenticated inference service. No prompt injection or secret exfiltration intent was found.
4
已掃描檔案
1,664
分析行數
15
複核項目
0
已忽略的誤報
已確認的安全問題 (2)
中
Unpinned Remote Package Execution
The SOP asks users or agents to run npx flow-nexus@latest and npx claude-flow@alpha. These dist-tags can execute package code that changes over time without a pinned version.
The cited commands directly invoke npm packages through mutable latest and alpha tags. This is a clear supply-chain risk when automated by an agent.
中
Unauthenticated Inference Service Exposure
The deployment example binds Uvicorn to all interfaces and maps port 8000 while exposing prediction and metrics endpoints without authentication.
The service binds to 0.0.0.0 and the deploy script publishes the port. The FastAPI example shows health, predict, and metrics routes with no access control.
能力複核項目 (15)
這些是真實的本機能力，對此技能而言可能是預期行為，因此需要複核，但不會按已確認的惡意行為計分。
中
Shell command substitution
npx claude-flow@alpha hooks session-restore --session-id "neural-training-$(date +%s)"
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"complete\", \"cluster_id\": \"[id]\", \"timestamp\": \"$(date -Iseconds)\"}"
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
CLUSTER_ID=$(npx claude-flow@alpha memory retrieve --key "neural/cluster-id" | jq -r '.value')
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"complete\", \"nodes\": 4, \"topology\": \"mesh\", \"timestamp\": \"$(date -I
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
CLUSTER_ID=$(npx claude-flow@alpha memory retrieve --key "neural/cluster-id" | jq -r '.value')
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
STATUS=$(npx claude-flow@alpha memory retrieve --key "neural/training-status" | jq -r '.value')
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"complete\", \"training_job\": \"$TRAINING_JOB_ID\", \"final_accuracy\": 0.94
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
CLUSTER_ID=$(npx claude-flow@alpha memory retrieve --key "neural/cluster-id" | jq -r '.value')
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
TRAINING_METRICS=$(npx claude-flow@alpha memory retrieve --key "neural/training-metrics" | jq -r '.v
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"inference_latency_ms\": 67, \"throughput_qps\": 1200, \"memory_mb\": 512, \"timestamp\":
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"complete\", \"validation_passed\": true, \"accuracy\": 0.94, \"latency_ms\":
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"model_id\": \"$TRAINING_JOB_ID\", \"template_published\": true, \"timestamp\": \"$(date
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"complete\", \"deployment\": \"docker\", \"api\": \"fastapi\", \"port\": 8000
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Shell command substitution
--value "{\"status\": \"success\", \"model_accuracy\": 0.94, \"latency_ms\": 67, \"deployed\": true,
This documented shell example invokes npx claude-flow@alpha while using command substitution. Running unpinned package commands from a skill can execute changing third-party code in the user environment.
中
Hardcoded IP address
CMD ["uvicorn", "serve:app", "--host", "0.0.0.0", "--port", "8000"]
The Dockerfile example binds Uvicorn to 0.0.0.0, and the deployment script maps port 8000. If copied as-is, it can expose an unauthenticated inference API.
風險因素
🔑 環境變數 (1)
⚙️ 外部命令 (87)
PROCESS.md:318 SKILL.md:8-35 SKILL.md:35-54 SKILL.md:54-60 SKILL.md:60-107 SKILL.md:107-113 SKILL.md:113-116 SKILL.md:116-136 SKILL.md:136-139 SKILL.md:139-166 SKILL.md:166-175 SKILL.md:175-180 SKILL.md:180-192 SKILL.md:192-251 SKILL.md:251-254 SKILL.md:254-295 SKILL.md:295-298 SKILL.md:298-335 SKILL.md:335-344 SKILL.md:344-348 SKILL.md:348-361 SKILL.md:361-384 SKILL.md:384-387 SKILL.md:387-421 SKILL.md:421-424 SKILL.md:424-471 SKILL.md:471-480 SKILL.md:480-484 SKILL.md:484-497 SKILL.md:497-566 SKILL.md:566-569 SKILL.md:569-598 SKILL.md:598-601 SKILL.md:601-642 SKILL.md:642-651 SKILL.md:651-655 SKILL.md:655-668 SKILL.md:668-706 SKILL.md:706-709 SKILL.md:709-729 SKILL.md:729-732 SKILL.md:732-854 SKILL.md:854-856 SKILL.md:856-859 SKILL.md:859-863 SKILL.md:863-866 SKILL.md:866-868 SKILL.md:868-874 SKILL.md:874-890 SKILL.md:890-900 SKILL.md:900-909 SKILL.md:909-951 SKILL.md:951-957 SKILL.md:112 SKILL.md:179 SKILL.md:194 SKILL.md:347 SKILL.md:363 SKILL.md:411 SKILL.md:459 SKILL.md:483 SKILL.md:499 SKILL.md:500 SKILL.md:579 SKILL.md:606 SKILL.md:654 SKILL.md:678 SKILL.md:728 SKILL.md:903 SKILL.md:908 SKILL.md:107-113 SKILL.md:175-180 SKILL.md:192-251 SKILL.md:344-348 SKILL.md:361-384 SKILL.md:387-421 SKILL.md:424-471 SKILL.md:480-484 SKILL.md:497-566 SKILL.md:569-598 SKILL.md:601-642 SKILL.md:651-655 SKILL.md:668-706 SKILL.md:709-729 SKILL.md:900-909 SKILL.md:457 SKILL.md:811
審計單位: codex 檢視審計紀錄 →
分享與引用此報告
分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。
複製報告連結
https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/audits/8?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown 徽章
[](https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural?utm_source=security_passport_badge)HTML 徽章
<a href="https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/security.svg" alt="Skillstore security assessment" loading="lazy"></a>嵌入卡片
<iframe src="https://skillstore.io/embed/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>學術引用 (APA · BibTeX · CFF)
APA 引用
DNYoussef. (2026). when-training-neural-networks-use-flow-nexus-neural security audit report (audit version 8) [Author version unspecified]. Skillstore. https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/audits/8BibTeX 引用
@techreport{dnyoussef-dnyoussef-when-training-neural-networks-use-flow-nexus-neural-2026,
author = {DNYoussef},
title = {when-training-neural-networks-use-flow-nexus-neural security audit report (audit version 8)},
institution = {Skillstore},
year = {2026},
number = {8},
url = {https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/audits/8},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "when-training-neural-networks-use-flow-nexus-neural security audit report (audit version 8)"
version: "unspecified"
type: report
authors:
- name: "DNYoussef"
date-released: "2026-07-05"
url: "https://skillstore.io/skills/dnyoussef-when-training-neural-networks-use-flow-nexus-neural/audits/8"
identifiers:
- type: other
value: "skillstore:dnyoussef-when-training-neural-networks-use-flow-nexus-neural:audit:8"
description: "Skillstore immutable audit report identifier"
Skillstore 評分
評分依據 證據置信度: 高 55
架構
85
可維護性
87
內容
72
社群
74
規範符合度
你可以打造什麼
規劃分散式訓練
建立包含架構、資料集、驗證與基準測試步驟的結構化訓練計畫。
協調 Flow Nexus 資源
對應叢集設定、節點部署、記憶體協調與分散式訓練作業。
準備模型服務
為已訓練模型草擬 Docker、FastAPI、健康檢查、指標與部署文件。
試試這些提示
規劃一次訓練執行
使用此技能，根據我的任務、資料集類型、目標準確率與部署目標，規劃一次 Flow Nexus 神經網路訓練執行。
設定架構
使用此技能，為我的神經網路提出模型架構、超參數、驗證檢查與 Flow Nexus 資源。
驗證與基準測試
使用此技能，為我的已訓練模型設計驗證測試、效能基準、驗收門檻與疑難排解步驟。
準備生產環境部署
使用此技能，為我的已訓練模型產出生產安全部署檢查清單、服務 API 計畫、監控計畫與推出步驟。
最佳實務
- 在執行命令前，替換所有佔位模型 ID、憑證、指標與資料集值。
- 鎖定套件版本，並在允許代理執行 shell 命令前先加以審查。
- 在公開推論服務前，加入驗證、存取控制與可觀測性。
避免
- 在共享或生產環境中原封不動地執行範例部署命令。
- 將示意性的準確率、延遲與吞吐量數字視為真實驗證證據。
- 在沒有驗證、速率限制、監控或復原計畫的情況下發布模型端點。
常見問題
此技能能協助什麼？
它有助於規劃並執行從設定到部署的 Flow Nexus 神經網路工作流程。
它會自行訓練模型嗎？
不會。它提供工作流程指引與範例，需要 Flow Nexus 工具和使用者提供的資料。
參考了哪些工具？
它參考 Flow Nexus MCP actions、Claude Flow hooks、E2B sandboxes、Docker、FastAPI 與 TensorFlow serving 概念。
我可以搭配 Claude Code 或 Codex 使用它嗎？
可以。它是為 Claude、Claude Code 與 Codex 相容工作流程撰寫的程序式技能。
執行命令前我應該審查什麼？
請審查套件版本、shell scripts、Docker 設定、網路暴露範圍、憑證與本機檔案路徑。
它適合用於生產環境部署嗎？
它可以支援生產環境規劃，但使用者必須加入安全控制、治理、測試與營運審查。
開發者詳細資訊
作者
DNYoussef授權
MIT
Skillstore 修訂
r1
版本提示
作者未聲明版本。
引用
c579efd02184f1128d6741d16e6c8d951d7d4208
維護新鮮度
2026/7/18
使用量
6 次下載 · 331 次瀏覽
檔案結構