POG Diagrams¶
This page contains all the visual diagrams for the Prompt Orchestration Governance framework.
Diagram 1: POG Dual Architecture Overview¶
This diagram illustrates the two core functions of POG: Prompt Warehouse Management and SDLC Integration.
graph TD
subgraph PW["Prompt Warehouse Management"]
PW1["Discover Prompts"]
PW2["Normalize & Structure"]
PW3["Validate & Test"]
PW4["Version & Store"]
PW5["Prompt Repository"]
end
subgraph SDLC["SDLC Integration Layer"]
S1[Requirements Phase]
S2[Design Phase]
S3[Development Phase]
S4[Testing Phase]
S5[Deployment Phase]
S6[Maintenance Phase]
end
PW1 --> PW2
PW2 --> PW3
PW3 --> PW4
PW4 --> PW5
PW5 -.provides prompts.-> S1
PW5 -.provides prompts.-> S2
PW5 -.provides prompts.-> S3
PW5 -.provides prompts.-> S4
PW5 -.provides prompts.-> S5
PW5 -.provides prompts.-> S6
S1 -.feedback & new prompts.-> PW1
S2 -.feedback & new prompts.-> PW1
S3 -.feedback & new prompts.-> PW1
S4 -.feedback & new prompts.-> PW1
S5 -.feedback & new prompts.-> PW1
S6 -.feedback & new prompts.-> PW1Description: The Prompt Warehouse Management function discovers, normalizes, validates, and stores prompts. The SDLC Integration Layer provides phase-specific prompts to each stage of software development. Feedback from SDLC usage continuously improves the prompt repository.
Diagram 2: SDLC Phase Prompt Flow¶
This diagram shows how different prompts are invoked at each SDLC phase with specific examples.
flowchart TD
subgraph Requirements["Requirements Phase"]
R1[User Story<br/>Elicitation]
R2[Acceptance<br/>Criteria Generator]
R3[Risk Analysis]
end
subgraph Design["Design Phase"]
D1[Architecture<br/>Pattern Advisor]
D2[API Design<br/>Assistant]
D3[Data Model<br/>Designer]
end
subgraph Development["Development Phase"]
DV1[Code<br/>Generator]
DV2[Code Review<br/>Assistant]
DV3[Documentation<br/>Generator]
end
subgraph Testing["Testing Phase"]
T1[Test Case<br/>Generator]
T2[Bug Analysis<br/>Assistant]
T3[Coverage<br/>Analyzer]
end
subgraph Deployment["Deployment Phase"]
DP1[Release Notes<br/>Generator]
DP2[Configuration<br/>Validator]
DP3[Rollback<br/>Planner]
end
subgraph Maintenance["Maintenance Phase"]
M1[Incident<br/>Analyzer]
M2[Performance<br/>Optimizer]
M3[Tech Debt<br/>Assessor]
end
Requirements ==> Design
Design ==> Development
Development ==> Testing
Testing ==> Deployment
Deployment ==> Maintenance
Maintenance -.continuous improvement.-> RequirementsDescription: Each SDLC phase has dedicated prompt categories that address specific needs. The cycle is continuous, with maintenance phase insights feeding back into requirements for the next iteration.
Diagram 3: Prompt Lifecycle State Machine¶
This diagram illustrates how prompts evolve from ad-hoc interactions to production-ready skill prompts.
stateDiagram-v2
[*] --> Interaction: Ad-hoc chat usage
Interaction --> Discovery: Identify value
Discovery --> Normalization: Extract & structure
Normalization --> Validation: Test & verify
Validation --> Repository: Approve & store
Repository --> Active: Deploy to production
Active --> Monitoring: Track usage
Monitoring --> Refinement: Collect feedback
Refinement --> Normalization: Improve
Monitoring --> Deprecated: Low usage/outdated
Deprecated --> [*]: Archive
Active --> [*]: End of lifecycle
note right of Interaction
Informal prompts used
in chat or exploration
end note
note right of Repository
Versioned, governed
production-ready prompts
end note
note right of Monitoring
Usage metrics, feedback,
effectiveness measurement
end noteDescription: Prompts begin as informal interactions, progress through discovery and normalization, get validated, and become production-ready skill prompts. Continuous monitoring enables refinement, and prompts can be deprecated when no longer useful.
Diagram 4: Meta-Loop - POG Self-Improvement¶
This diagram shows how POG continuously improves its own prompt library through feedback and usage analysis.
graph TD
A[POG Prompt Repository] --> B[Projects Use Prompts]
B --> C[Collect Usage Data]
C --> D[Analyze Metrics & Feedback]
D --> E{Evaluation}
E -->|High value,<br/>frequently used| F[Promote to<br/>Core Library]
E -->|Effective but<br/>needs refinement| G[Improve<br/>Existing Prompt]
E -->|Gap identified| H[Create<br/>New Prompt]
E -->|Low usage,<br/>outdated| I[Deprecate<br/>or Archive]
F --> J[Update Repository]
G --> J
H --> J
I --> K[Archive]
J --> A
B -.new prompts from projects.-> L[Discovery Queue]
L --> M[Review & Normalize]
M --> JDescription: POG implements a meta-loop where prompt usage is continuously monitored, analyzed, and fed back into the repository. High-value prompts are promoted, existing ones are refined, gaps trigger new prompt creation, and low-value prompts are deprecated. This ensures the prompt library evolves with organizational needs.
Diagram 5: Orchestration Layers Hierarchy¶
This diagram shows the hierarchical layers that organize prompts by scope and context.
graph TB
A[Runtime Context Layer]
B[Task Layer]
C[Domain Layer]
D[Foundation Layer]
A --> B
B --> C
C --> D
A1["`**User Input**
Project context
Environment variables
Session state`"]
B1["`**Specific Tasks**
Generate unit tests
Review API design
Create user stories
Analyze incident`"]
C1["`**Domain Rules**
Healthcare regulations
Financial compliance
E-commerce patterns
IoT protocols`"]
D1["`**Core Capabilities**
Safety constraints
Output formatting
Error handling
Security policies`"]
A1 -.describes.-> A
B1 -.describes.-> B
C1 -.describes.-> C
D1 -.describes.-> DDescription: Prompts are organized in hierarchical layers, from the foundational system capabilities up to runtime-specific context. Each layer builds upon the layers below it:
- Foundation Layer: Universal system policies and constraints
- Domain Layer: Business-specific rules and patterns
- Task Layer: Specific development tasks and intents
- Runtime Context Layer: Dynamic, session-specific information
This layering enables prompt composition and context management.
Diagram 6: Prompt Lifecycle Flow (Detailed)¶
This diagram provides a detailed view of the prompt management lifecycle with decision points.
flowchart TD
Start([New Prompt Need]) --> Source{Source?}
Source -->|Chat interaction| Chat[Interaction Prompt]
Source -->|Project discovery| Project[Project Prompt]
Source -->|Team contribution| Contrib[Contributed Prompt]
Chat --> Capture[Capture Prompt]
Project --> Capture
Contrib --> Capture
Capture --> Review{Worth<br/>formalizing?}
Review -->|No| Discard[Discard]
Review -->|Yes| Normalize[Normalize<br/>& Structure]
Normalize --> Param[Parameterize<br/>Variables]
Param --> Meta[Add Metadata]
Meta --> Eval[Create<br/>Evaluation Cases]
Eval --> Test[Test Prompt]
Test --> Pass{Tests<br/>pass?}
Pass -->|No| Fix[Refine Prompt]
Fix --> Test
Pass -->|Yes| Version[Assign Version]
Version --> Cat[Categorize by<br/>SDLC Phase]
Cat --> Store[Store in<br/>Repository]
Store --> Deploy[Deploy to<br/>Production]
Deploy --> Monitor[Monitor Usage]
Monitor --> Feedback{Feedback?}
Feedback -->|Needs improvement| Fix
Feedback -->|Working well| Continue[Continue<br/>Monitoring]
Feedback -->|Not used| Archive[Archive]
Continue --> Monitor
Discard --> End([End])
Archive --> EndDescription: This detailed flow shows the complete journey of a prompt from identification through deployment and monitoring, including quality gates and decision points that ensure only valuable, tested prompts reach production.
For more information, see the main whitepaper.