Sean Michael Dugan
ALL PROJECTS
TOOLING / HOW I SHIP

Agent-Assisted Delivery Pipeline

A short look at how I build and ship: the harnesses I use, the checks, and how releases go out.

Cursor · Cursor cloud agents · Codex · Grok Bot agents · AGENTS.md · release-kit · AWS

How work flows

11+ years of building software, combined with newer AI development tools. Agents do most of the drafting, checks run locally, and I approve every merge and release.

From request to live
Idea or request
Agents draftCursor · Codex
Pull request
Local checksBuild, tests, privacy, infra checks
I approve?
Changes → back to the agents
ReleaseFrom my machine
Live check

Harnesses I use

Who does what
CursorLocalEditor on my own machine.
Cursor cloud agents & ProjectsCloudA coordinator splits work and agents open PRs.
CodexFeature workLonger feature work on HC Survive.
Grok Bot agentsAssistantsPlanning, deploys and tracking work, each with its own role.
AGENTS.mdOne per repo. Shared rules every harness reads first.

Releasing

This site is static on AWS, released from my machine after a dry run, with no hosted CI spend.

For other projects I wrote release-kit (rk), a small CLI. Each project keeps its own config, and the steps stay the same for every project.

release-kit · one config per project, same steps everywhere
release-kit.config.json
  • deployType
  • Package steps
  • Pointer to the live release
Separate, ignored fileAccount details
rk · same for every project
package
plan
approve
release
roll back

rk plan prints a digest. Nothing is written until I pass it back with --approve.

  • Static sites are supported today. Terraform and containers are planned under the same steps. Planned
  • First used to ship HC Survive.

Asset pipeline In progress

Codex is adding 3D models to HC Survive.

How the game loads a model
Small low-poly GLBPer model
Load
DrawRepeated models, like trees, in batches
If it fails: built-in shapes

In early tests, batching cut draw calls sharply. Frame rate still needs testing on real hardware.

HC Survivehcsurvive.io ↗ (opens in a new tab)

Hardware

Part of the pipeline runs on a custom, dedicated headless machine of my own. Background jobs are handed off to it, so they run on my own hardware instead of in the cloud.

NEXT PROJECTLocal Discovery Platform