when-chaining-agent-pipelines-use-stream-chain
以串流處理串聯代理流程
建構複雜的多代理工作流程，讓每個代理的輸出無縫流向下一個。Stream-chain 透過即時資料流監控，協調順序與平行管線。
Die Skill-ZIP herunterladen
In Claude hochladen
Gehe zu Einstellungen → Fähigkeiten → Skills → Skill hochladen
Einschalten und loslegen
Teste es
Verwendung von "when-chaining-agent-pipelines-use-stream-chain". Design a research pipeline for market analysis with 5 stages
Erwartetes Ergebnis:
- Pipeline Design: research, data-collection, analysis, synthesis, reporting
- Stage 1 (Research): Defines scope and objectives
- Stage 2 (Data Collection): Gathers market data via parallel agents
- Stage 3 (Analysis): Processes data with 3 parallel analyzers
- Stage 4 (Synthesis): Combines analysis results
- Stage 5 (Reporting): Creates final market report
- Throughput: 12 items per minute, Error rate: 1.2 percent, Total latency: 4.2 minutes
Verwendung von "when-chaining-agent-pipelines-use-stream-chain". Create a CI/CD pipeline that builds, tests, and deploys in parallel
Erwartetes Ergebnis:
- Pipeline Structure: build stage with parallel test and deploy branches
- Build Stage: Compiles code and generates artifacts
- Parallel Branch A: Runs unit tests, integration tests, security scans
- Parallel Branch B: Prepares deployment configurations
- Merge Point: Combines test results with deployment readiness
- Deploy Stage: Pushes to staging or production environment
Sicherheitsaudit
SicherAll 79 static findings are false positives. The skill contains only documentation describing Claude Flow pipeline orchestration patterns. No executable code, network calls, or file system access exist. The 'Weak cryptographic algorithm' flags are triggered by documentation text (e.g., 'stage', 'spawn', 'shell'). The 'external_commands' flags are markdown code formatting backticks, not shell execution.
Risikofaktoren
⚡ Enthält Skripte
🌐 Netzwerkzugriff
📁 Dateisystemzugriff
🔑 Umgebungsvariablen
⚙️ Externe Befehle
Qualitätsbewertung
Was du bauen kannst
多階段程式碼審查
串聯研究、分析、撰寫、測試與審查代理，進行系統化的程式碼開發，並在每個階段驗證輸出。
平行資料分析
以平行分析階段處理大型資料集，並彙整串流結果與即時監控。
自動化內容管線
在順序管線中協調研究、撰寫、編輯與發佈代理，以維持一致的內容品質。
Probiere diese Prompts
Create a 3-stage sequential pipeline. Stage 1 researches a topic, Stage 2 analyzes findings, Stage 3 compiles a report. Use memory coordination for data passing.
Design a pipeline where initial data splits into 3 parallel analysis branches that converge into a final synthesis stage.
Build a hybrid pipeline. Sequential stages 1-2, then parallel stages 3a-3b-3c converging to stage 4. Include throughput monitoring every 30 seconds.
Create a 4-stage pipeline with error handling. Each stage validates input, retries on failure (max 3 attempts), and stores error metrics.
Bewährte Verfahren
- 設計清楚的階段邊界，單一職責，並使用已驗證的資料結構
- 實作背壓監控，防止在階段邊界出現佇列溢位
- 使用記憶體協調器進行狀態持久化，並在失敗時允許階段重啟
Vermeiden
- 避免在管線階段之間建立會導致無限迴圈的循環相依
- 不要硬編碼階段設定，使用外部設計檔以提升彈性
- 切勿略過階段間驗證，未檢查的資料毀損會在管線中擴散