| name | pm-agent-orchestration |
| description | Orchestrate PM-Agent workflows for case management, task delegation, and artifact pipeline coordination. Use when managing multi-step projects, delegating tasks to sub-agents, or tracking work through CMMN stages. |
| license | Complete terms in LICENSE.txt |
PM-Agent Orchestration
Coordinate project work through PM-Agent - the SEA-Forge™ orchestration layer for case management, task delegation, and artifact pipelines.
For reference: Project/Case Management Handbook | Case Management Skill
When to Use This Skill
Use pm-agent-orchestration when:
- Managing Complex Projects: Work spanning multiple phases, teams, or dependencies
- Delegating to Sub-Agents: Breaking work into tasks for specialized agents
- Tracking Artifact Pipelines: Moving artifacts through CMMN stages
- Coordinating Handoffs: Ensuring smooth transitions between stages
- Reporting Progress: Generating status updates across cases
PM-Agent Architecture
┌─────────────────────────────────────────────────────────────┐
│ PM-Agent │
│ ┌─────────────┐ ┌──────────────┐ ┌────────────────────┐ │
│ │ Case Manager │ │ Task Router │ │ Artifact Pipeline │ │
│ └─────────────┘ └──────────────┘ └────────────────────┘ │
│ │ │ │ │
│ ▼ ▼ ▼ │
│ ┌─────────────┐ ┌──────────────┐ ┌────────────────────┐ │
│ │ NATS Bus │◄─┤ Sub-Agents │ │ Knowledge Graph │ │
│ └─────────────┘ └──────────────┘ └────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
Core Workflows
1. Case Creation
Create a new case for a project or initiative:
command: pm.case.create
payload:
title: "SEA-Forge Skills Optimization"
description: "Optimize skills, handbooks, and documentation"
desiredOutcome: "All skills whitelabeled, new artifacts added, docs aligned"
timeline:
startDate: "2026-01-06"
targetDate: "2026-01-20"
semanticRefs:
- conceptId: "sea:SkillDefinition"
- conceptId: "sea:Documentation"
Response:
caseId: "case-2026-001"
status: "created"
stage: "intake"
2. Task Breakdown
Break a case into delegatable tasks:
command: pm.task.plan
payload:
caseId: "case-2026-001"
breakdown:
strategy: "vertical-slice" # or "horizontal", "milestone"
maxParallelism: 3
Generated Tasks:
tasks:
- id: "task-001"
name: "Whitelabel docx skill"
stage: "execution"
agent: "code-agent"
dependencies: []
- id: "task-002"
name: "Add Notebook artifact type"
stage: "execution"
agent: "code-agent"
dependencies: []
- id: "task-003"
name: "Create semantic-anchoring skill"
stage: "execution"
agent: "code-agent"
dependencies: ["task-001", "task-002"]
3. Task Delegation
Delegate tasks to specialized sub-agents:
command: pm.task.delegate
payload:
taskId: "task-001"
agent: "code-agent"
context:
files:
- ".github/skills/docx/SKILL.md"
- ".github/skills/docx/scripts/document.py"
instructions: "Replace Claude references with SEA-Forge Agent"
successCriteria:
- "No Claude references remain"
- "All tests pass"
Agent Types:
| Agent | Specialization |
|---|---|
code-agent |
Code generation, refactoring |
review-agent |
Code review, PR analysis |
test-agent |
Test generation, verification |
doc-agent |
Documentation, handbooks |
research-agent |
Codebase exploration, analysis |
4. Progress Tracking
Monitor case and task progress:
command: pm.case.status
payload:
caseId: "case-2026-001"
Response:
case:
id: "case-2026-001"
stage: "execution"
progress:
completed: 5
inProgress: 2
pending: 3
total: 10
percentage: 50
timeline:
onTrack: true
estimatedCompletion: "2026-01-18"
blockers: []
recentActivity:
- "task-001 completed: docx skill whitelabeled"
- "task-002 completed: Notebook artifact added"
5. Artifact Pipeline
Move artifacts through CMMN stages:
command: pm.artifact.advance
payload:
artifactId: "artifact-2026-001"
fromStage: "draft"
toStage: "review"
reviewer: "human" # or "agent", "automated"
Artifact Stages:
draft → review → approved → published → archived
↓
rejected → draft (revision)
Event Integration (NATS)
PM-Agent publishes events for observability:
# Case created
subject: "sea.pm.case.created"
payload:
caseId: "case-2026-001"
timestamp: "2026-01-06T12:00:00Z"
# Task completed
subject: "sea.pm.task.completed"
payload:
taskId: "task-001"
caseId: "case-2026-001"
duration: "PT2H30M"
result: "success"
# Stage transition
subject: "sea.pm.artifact.stage.changed"
payload:
artifactId: "artifact-2026-001"
from: "draft"
to: "review"
Delegation Patterns
Sequential Execution
Tasks run one after another:
execution:
strategy: "sequential"
tasks:
- "task-001" # Runs first
- "task-002" # Runs after task-001
- "task-003" # Runs after task-002
Parallel Execution
Independent tasks run concurrently:
execution:
strategy: "parallel"
maxConcurrency: 3
tasks:
- "task-001" # All run
- "task-002" # simultaneously
- "task-003" # up to max
DAG Execution
Tasks with dependencies:
execution:
strategy: "dag"
tasks:
- id: "task-001"
dependencies: []
- id: "task-002"
dependencies: []
- id: "task-003"
dependencies: ["task-001", "task-002"] # Waits for both
Quality Guidelines
1. Clear Success Criteria
Every task must have measurable success criteria:
successCriteria:
- metric: "tests_passing"
threshold: 100
unit: "percent"
- metric: "lint_errors"
threshold: 0
unit: "count"
2. Context Minimization
Provide only relevant context to sub-agents:
# ❌ Too much context
context:
files: ["**/*"] # Entire repo
# ✅ Focused context
context:
files:
- ".github/skills/docx/SKILL.md"
- ".github/skills/docx/scripts/document.py"
3. Atomic Tasks
Break work into atomic, verifiable units:
# ❌ Too broad
tasks:
- name: "Optimize all skills"
# ✅ Atomic
tasks:
- name: "Whitelabel docx skill"
- name: "Whitelabel skill-creator skill"
- name: "Add Notebook artifact type"