case-management

Case Management (CMMN)

Orchestrate knowledge-intensive work using SEA-Forge™'s case-based approach. Cases adapt to reality rather than forcing reality into rigid plans, making them ideal for research, consulting, creative projects, and complex discretionary workflows.

For reference: Case Management Handbook | SDS-012: Case Management Service

---

When to Use This Skill

Use case-management when:

1. Discretionary Work: Outcome is clear but path is uncertain (research, design)
2. Knowledge Reification: Raw notes need to become formal specifications or assets
3. Hybrid Teams: Coordinating mix of human experts and AI specialists
4. Complex Workflows: Multiple stages with event-driven transitions
5. Asset Formation: Work products should become reusable organizational capital

---

Core Concepts

CMMN Primitives

Primitive	Purpose	Example
Case	Container for knowledge work	"Market Research Q1", "Feature X Design"
Stage	Logical grouping of tasks	"Discovery", "Analysis", "Recommendations"
Task	Unit of work (discretionary or mandatory)	"Interview stakeholders", "Write spec"
Sentry	Event-driven conditional trigger	"When 3 interviews complete, activate analysis"
Milestone	Observable achievement	"Discovery complete", "Spec approved"

Artifact Pipeline (Four Reification Stages)

Cognitive Artifact → Intellectual Artifact → Information Product → Intellectual Capital
(notes, sketches) → (specs, designs)      → (software, reports) → (organizational assets)

Key Principle: Artifacts must progress through stages in order. No "teleportation" – every transition requires provenance via TransitionTokens.

PM-Agent Orchestration

The Project Manager Agent (PM-Agent) serves as conductor for hybrid human-AI teams:

• Activates discretionary tasks when sentries fire
• Assigns tasks to appropriate specialists (human or AI)
• Tracks artifact pipeline progression
• Enforces governance (Sovereign Gates for high-impact decisions)

---

Workflow: Creating a Case

Step 1: Define Desired Outcome

Every case starts with a clear goal:

Template:

case:
  id: "case-2026-001"
  title: "<Brief descriptive title>"
  desiredOutcome: "<What success looks like>"
  context: "<Why this case exists>"
  stakeholders:
    - name: "<Stakeholder>"
      role: "<Their role>"

Example:

case:
  id: "case-2026-001"
  title: "Optimize Skills Architecture"
  desiredOutcome: "All skills whitelabeled, cognitive-artifacts-builder created, handbooks aligned"
  context: "Current skills reference Claude; need SEA-Forge branding and CADSL integration"
  stakeholders:
    - name: "Architecture Team"
      role: "Owner"
    - name: "Documentation Team"
      role: "Contributor"

Step 2: Define Stages

Break case into logical phases:

Stage Types:

• Discretionary: Can be skipped if not needed
• Mandatory: Must be completed
• Repeatable: Can occur multiple times

Template:

stages:
  - id: "<stage-id>"
    name: "<Stage Name>"
    type: "discretionary|mandatory|repeatable"
    entryCriteria:  # Sentry conditions
      - "<condition>"
    exitCriteria:   # Milestones
      - "<achievement>"

Example:

stages:
  - id: "analysis"
    name: "Current State Analysis"
    type: "mandatory"
    entryCriteria:
      - "Case activated"
    exitCriteria:
      - "All skills inventoried"
      - "Brand references identified"

  - id: "whitelabeling"
    name: "Whitelabeling"
    type: "mandatory"
    entryCriteria:
      - "Analysis complete"
    exitCriteria:
      - "5 skills whitelabeled"

  - id: "transformation"
    name: "Transformation"
    type: "mandatory"
    entryCriteria:
      - "Whitelabeling complete"
    exitCriteria:
      - "cognitive-artifacts-builder created"
      - "CADSL examples included"

Step 3: Define Tasks

Tasks are the atomic units of work:

Task Types:

• Human Task: Requires human judgment
• Process Task: Can be automated
• Case Task: Nested sub-case

Template:

tasks:
  - id: "<task-id>"
    name: "<Task name>"
    type: "human|process|case"
    stage: "<stage-id>"
    assignedTo: "<member-id>"
    isDiscretionary: true|false
    activationSentry:
      conditions:
        - "<condition>"

Example:

tasks:
  - id: "task-001"
    name: "Inventory all skills"
    type: "process"
    stage: "analysis"
    assignedTo: "agent-specialist-analysis"
    isDiscretionary: false

  - id: "task-002"
    name: "Identify Claude references"
    type: "process"
    stage: "analysis"
    assignedTo: "agent-specialist-analysis"
    isDiscretionary: false
    activationSentry:
      conditions:
        - "task-001 completed"

  - id: "task-003"
    name: "Whitelabel web-artifacts-builder"
    type: "process"
    stage: "whitelabeling"
    assignedTo: "agent-specialist-refactor"
    isDiscretionary: false
    activationSentry:
      conditions:
        - "analysis stage complete"

Step 4: Configure PM-Agent

PM-Agent manages case lifecycle:

Configuration:

pmAgent:
  caseId: "case-2026-001"
  capabilities:
    - taskActivation: true
    - artifactTracking: true
    - sovereignGateEnforcement: true

  rules:
    # Activate discretionary tasks when conditions met
    - trigger: "sentry-fired"
      action: "activate-task"

    # Track artifact pipeline transitions
    - trigger: "artifact-created"
      action: "record-transition-token"

    # Require human approval for capital formation
    - trigger: "capitalization-flow"
      action: "request-sovereign-signature"

Step 5: Track Artifacts

As work progresses, artifacts move through pipeline:

Artifact Tracking:

artifacts:
  - id: "artifact-001"
    name: "Skills Analysis Report"
    stage: "IntellectualArtifact"  # Formal spec
    case: "case-2026-001"
    createdBy: "agent-specialist-analysis"
    provenance:
      - transition: "CognitiveArtifact → IntellectualArtifact"
        transitionToken: "token-001"
        timestamp: "2026-01-06T10:00:00Z"

  - id: "artifact-002"
    name: "cognitive-artifacts-builder Skill"
    stage: "InformationProduct"  # Executable work product
    case: "case-2026-001"
    createdBy: "agent-specialist-refactor"
    provenance:
      - transition: "IntellectualArtifact → InformationProduct"
        transitionToken: "token-002"
        timestamp: "2026-01-06T12:00:00Z"

---

Common Case Patterns

Research Case

Stages: Literature Review → Hypothesis Formation → Experiments → Analysis → Publication

Key Sentries:

• Activate experiments when 10+ papers reviewed
• Activate analysis when 3+ experiments complete
• Activate publication when findings validated

Artifacts:

• Literature notes (Cognitive) → Research design (Intellectual) → Paper draft (Product) → Published paper (Capital)

Development Case (Spec-First)

Stages: Ideation → Specification → Implementation → Testing → Deployment

Key Sentries:

• Activate implementation when ADR + PRD + SDS approved
• Activate testing when code generation complete
• Activate deployment when all tests pass

Artifacts:

• Design notes (Cognitive) → ADR/PRD/SDS (Intellectual) → Generated code (Product) → Released feature (Capital)

Consulting Case

Stages: Discovery → Analysis → Recommendations → Presentation → Handoff

Key Sentries:

• Activate analysis when 5+ stakeholder interviews complete
• Activate recommendations when analysis validated
• Activate presentation when recommendations approved

Artifacts:

• Interview notes (Cognitive) → Analysis doc (Intellectual) → Recommendation deck (Product) → Client asset (Capital)

---

Integration Points

Cognitive Artifacts (CADSL)

Cases can use cognitive artifacts (checklists, decision trees, kanban) for visualization:

case:
  id: "case-2026-001"
  artifacts:
    - type: "Kanban"
      artifactId: "kanban-001"
      columns: ["Backlog", "In Progress", "Review", "Done"]

    - type: "Checklist"
      artifactId: "checklist-001"
      items: ["Whitelabel skills", "Create new skills", "Update handbooks"]

Use cognitive-artifacts-builder skill to generate these visualizations.

Knowledge Graph (DomainForge™)

Link cases and artifacts to semantic concepts:

case:
  semanticRefs:
    - conceptId: "sea:KnowledgeWork"
    - conceptId: "sea:DiscretionaryWorkflow"

artifacts:
  - id: "artifact-001"
    semanticRefs:
      - conceptId: "sea:Specification"
      - conceptId: "sea:IntellectualArtifact"

Internal Federated Ledger (IFL)

Every case state transition creates immutable TransitionToken:

transitionTokens:
  - id: "token-001"
    case: "case-2026-001"
    transition: "Created → Active"
    timestamp: "2026-01-06T09:00:00Z"
    ifl:hash: "sha256:abc123..."

  - id: "token-002"
    case: "case-2026-001"
    transition: "Stage[analysis] → Complete"
    timestamp: "2026-01-06T11:00:00Z"
    ifl:hash: "sha256:def456..."

No Teleportation Policy: Artifacts cannot skip stages. Every transition must have provenance token.

Observability (OpenTelemetry)

Track case metrics:

cmmn_cases_active_total{type="research"}
cmmn_stages_completed_total{case_id="case-2026-001"}
cmmn_artifacts_reified_total{stage="IntellectualCapital"}
cmmn_sovereign_gates_triggered_total

---

Governance Policies

Single Agent Control

Policy: AI members must be managed exclusively by PM-Agent

Enforcement: Case members cannot self-assign tasks or activate stages

No Teleportation

Policy: Artifacts must progress through pipeline in order

Example Violation:

# ❌ VIOLATION: Skipping IntellectualArtifact stage
artifact:
  stage: "InformationProduct"
  provenance: []  # No IntellectualArtifact transition token

Valid:

# ✅ VALID: Complete provenance chain
artifact:
  stage: "InformationProduct"
  provenance:
    - transition: "CognitiveArtifact → IntellectualArtifact"
      token: "token-001"
    - transition: "IntellectualArtifact → InformationProduct"
      token: "token-002"

Sovereign Gate

Policy: High-impact capital conversions require human signature

Trigger: When artifact transitions to IntellectualCapital

Action: PM-Agent requests SignatureToken from human sovereign

---

PM-Agent Configuration

Role Assignment

PM-Agent assigns tasks based on capability matching:

members:
  - id: "agent-specialist-analysis"
    type: "ai"
    capabilities: ["analysis", "research", "synthesis"]

  - id: "agent-specialist-refactor"
    type: "ai"
    capabilities: ["code-generation", "refactoring"]

  - id: "human-architect"
    type: "human"
    capabilities: ["architecture-decisions", "sovereign-approval"]

taskAssignment:
  rules:
    - taskType: "analysis"
      assignTo: "agent-specialist-analysis"

    - taskType: "refactoring"
      assignTo: "agent-specialist-refactor"

    - taskType: "sovereign-gate"
      assignTo: "human-architect"

Sentry Evaluation

PM-Agent continuously evaluates sentries:

sentries:
  - id: "activate-analysis"
    conditions:
      - "task-001.status == 'complete'"
      - "task-002.status == 'complete'"
    action:
      - "activate-stage('analysis')"
      - "notify-members(['agent-specialist-analysis'])"

---

Examples by Domain

Software Development

case:
  title: "Implement CADSL Renderer"
  stages:
    - Ideation (brainstorm, spike)
    - Specification (ADR, SDS, SEA™)
    - Implementation (codegen)
    - Testing (unit, integration)
    - Deployment (CI/CD)

Research Project

case:
  title: "LLM as Idiolect Study"
  stages:
    - Literature Review
    - Hypothesis Formation
    - Experiment Design
    - Data Collection
    - Analysis
    - Publication

Consulting Engagement

case:
  title: "Digital Transformation Roadmap"
  stages:
    - Discovery (stakeholder interviews)
    - Current State Analysis
    - Future State Design
    - Recommendations
    - Presentation

---

Quality Guidelines

1. Clear Desired Outcomes

❌ Bad: Vague or unachievable

desiredOutcome: "Make things better"

✅ Good: Specific and measurable

desiredOutcome: "All 17 skills whitelabeled with SEA-Forge branding; 3 new skills created; handbooks aligned with skills"

2. Proper Sentry Design

❌ Bad: Missing conditions

sentry:
  action: "activate-stage('analysis')"
  # No conditions specified

✅ Good: Explicit conditions

sentry:
  conditions:
    - "task-inventory.status == 'complete'"
    - "task-analysis.status == 'complete'"
  action: "activate-stage('whitelabeling')"

3. Complete Provenance

Always track artifact transitions:

artifact:
  id: "spec-001"
  provenance:
    - stage: "CognitiveArtifact"
      timestamp: "2026-01-05T14:00:00Z"
      token: "token-001"
    - stage: "IntellectualArtifact"
      timestamp: "2026-01-06T10:00:00Z"
      token: "token-002"

---

References

• Case Management Handbook
• SDS-012: Case Management Service
• CMMN Reference Guide
• Artifact Pipeline Management
• PM-Agent Orchestration