Creating & Managing HarnessesVirtual

Automate Without Losing Control

Advanced patterns for creating automated agents, harnesses, and governed orchestration systems.

An advanced workshop for teams ready to create automated agents and production-grade harnesses. You will learn agent architecture patterns, runtime and identity models, skills and subagents, state management, orchestration pipelines, testing, security, telemetry, evaluation, PR review automation, CI/CD integration, SDK usage, plugin development, and the governance needed to run agents responsibly.

Investment per seat
$299
Format
Live and interactive intensive
Reserve a seat →Train your team →

Workshop 3 of 3 in the Accelerator

Who this workshop is for

  • Senior Engineers
  • Architects
  • Platform Teams
  • DevOps/SRE
  • QA Automation Leads
  • Security/AppSec
  • Engineering Leaders Creating Automated Agents

Why it matters

Choose this if your technical teams are ready to move beyond one-off prompting into governed agents, harnesses, evals, and automation.

Leave with

A working model for creating automated agents and harnesses that can be tested, monitored, governed, integrated into delivery pipelines, and improved over time.

Schedule

Upcoming sessions

Sep 2, 2026Wed · 12 PM – 5 PM ET · Virtual
Reserve a seat →
Oct 7, 2026Wed · 12 PM – 5 PM ET · Virtual
Reserve a seat →
Dec 3, 2026Thu · 12 PM – 5 PM ET · Virtual
Reserve a seat →
Private workshopsVirtual or in person · Teams of 20+ · Customizable
Book a team workshop consult →

Harness architecture

Create automated agents with control, telemetry, and trust boundaries.

This workshop turns orchestration patterns into a working harness model: runtime, identity, registry, tools, state, evaluation, telemetry, and human escalation.

RUNTIME
Run agents inside clear execution boundaries
IDENTITY
Scope credentials and permissions per job type
REGISTRY
Share skills, subagents, tools, and workflows
TELEMETRY
Emit structured logs that improve future agent behavior
GATES
Use risk, score, review, and approval logic before action

Outcomes

What you'll take back to work

Design harness architectures using runtime, identity, registry, and orchestration patterns

Build modular skills and subagents that compose into larger automated workflows

Implement plan, risk, score, and action loops with clear human escalation thresholds

Test, observe, and benchmark agent behavior with regression suites and telemetry

Integrate agents into PR review, CI/CD, webhooks, scheduled jobs, SDKs, and plugin systems

Apply security, trust boundary, governance, and cost controls to production agent workflows

Prereqs

You should be comfortable with agent-ready specs, coding-agent workflows, and structured AI-assisted delivery. This workshop assumes you are ready to design automated agents and harnesses, not just write prompts.

Curriculum

What we cover

Practical, hands-on modules built from what we run with clients.

Agent Architecture Patterns

  • Orchestrator/worker patterns, parallel agents, and sequential pipelines
  • The runtime / harness / identity / registry model as a transferable architectural pattern
  • How these components compose into a working system

OpenCode/PI

  • Setup and configuration
  • Core concepts and use cases
  • Integration with your workflow

Skills and Subagents

  • Building modular, reusable skills
  • Designing subagents for specific tasks
  • Composing skills into larger workflows

State Management

  • Handling memory and session context
  • Managing state across long-running workflows
  • Context persistence strategies

Identifying Common Tasks for the Agent

  • Task analysis and decomposition using the Tier 1 / Tier 2 taxonomy
  • Patterns worth automating and how to prioritize them
  • Building a task inventory for your team

The Plan > Risk > Score > Action Loop

  • Designing agents that assess a plan before acting
  • Scoring confidence and defining thresholds for autonomous action
  • When to act, when to defer to an operator, and how to implement that decision

Automating Orchestration

  • Building orchestration pipelines
  • Multi-model workflows: routing tasks between models
  • Scheduling and event-driven triggers

QA Orchestration: End-to-End Case Study

  • A fully worked example: trigger > research > plan > risk > score > action > results > gate > iterate
  • How to apply this pattern to your own domain
  • Capstone reference architecture

Testing Agents

  • Unit testing prompts
  • Integration testing pipelines
  • Regression suites for agent behavior

Error Handling and Retries

  • Graceful degradation strategies
  • Fallback logic and retry patterns
  • Logging and surfacing failures

Security and Trust Boundaries

  • Prompt injection risks and mitigations
  • Identity scoping: creating identities per runtime and runtimes per job type
  • Least-privilege tool access and credential handling

Governance and Responsible AI

  • IP, data privacy, and disclosure obligations
  • Organizational policy and compliance
  • Human escalation protocols: designing agents that know when to stop

Cost Management

  • Token budgeting and caching strategies
  • Model selection tradeoffs at scale
  • Monitoring and alerting on spend

Observability and Telemetry as an Active Agent Input

  • Every harness should emit structured logs: sessions, token usage, specs, and work performed
  • Downstream agents querying telemetry to actively improve workflows
  • Identifying context gaps, surfacing process reports, and closing the improvement loop

Evaluation and Benchmarking

  • Measuring agent output quality over time
  • Regression detection
  • Building evaluation datasets

PR Reviews

  • Automating PR review workflows
  • Integrating agents into code review processes
  • Quality gates and approval logic

Deployment and CI/CD Integration

  • Running agents in pipelines
  • Webhooks and scheduled jobs
  • Environment management and Tier 2 deployment patterns

Claude SDK

  • SDK overview and setup
  • Building custom integrations
  • Advanced API usage patterns

Plugin Development

  • Designing and building plugins
  • Plugin architecture and packaging
  • Publishing and maintaining plugins

Capstone Assessment

  • Build and demo a working harness or agent pipeline
  • Peer or instructor review
  • Certification awarded upon completion

Community and Continued Learning

  • Where to go after certification
  • Staying current as models evolve
  • Contributing back to the community

Your instructors

Taught by practitioners

Elliott Fouts

Elliott Fouts

CTO

This Dot Labs

Elliott leads technical direction at This Dot Labs, with a focus on helping engineering teams adopt AI tooling that actually ships. He brings 25+ years in engineering across production AI systems, Claude Code workflows, technical architecture, and team AI adoption.

Jonathan Fontanez

Jonathan Fontanez

Engineering Lead, AI

This Dot Labs

Jonathan is an Engineering Lead, AI at This Dot Labs with 20 years of engineering experience helping teams design, build, and operationalize production software.

Rob Ocel

Rob Ocel

VP, Innovation

This Dot Labs

Rob is VP, Innovation at This Dot Labs with 20 years of experience helping teams adopt emerging technologies and turn new ideas into practical software delivery practices.

Keep exploring

Continue down the AI acceleration path

1/3 · FoundationsAI Is Already in Your SDLCFoundations for context engineering and accelerated AI utilization across software delivery.2-hour interactive live training · $199 per participant2/3 · Planning Work & SpecificationsMake the Work Agent-ReadyPlanning work, specifications, and handoffs for people using coding agents.Live and interactive intensive · $299 per participant

Need the full path?

The Agentic AI Software Delivery Lifecycle Accelerator Program.

Combine all three workshops to accelerate your organization through one connected progression: Foundations, Planning Work & Specifications, and Creating & Managing Harnesses.

Ask about team rollout →

Get certified

Become CAWA Certified.

Complete the full Agentic SDLC Accelerator and earn a practical credential in AI-assisted delivery.

Explore certification →

Tailored & custom workshops

Build a private workshop around your team, tools, and delivery process.

For teams of 20 or more, we can customize the syllabus around your SDLC, current AI tools, codebase reality, product process, governance concerns, and adoption goals.

Get private workshop recommendation →

Get a recommendation

Tell us what your team is trying to change.

Share your team size, current AI usage, workflow pain, and timing. We'll recommend the fastest workshop path: a public seat, a focused private workshop, or the full accelerator.

What happens next

We'll review your goals and reply with the best workshop path, suggested audience, and next steps for scheduling.

Prefer email? Reach us directly at workshops@thisdot.co.