Agent PMO

Stop Watching One Agent. Start Running Twenty.

🌐 Website: nimblesite.github.io/AgentPMOWorkflow

flowchart TD
    D["PMO Dashboard · scans every repo every 3 min"]
    D --> A1["✓ app-backend · PR #47 · CI green"]
    D --> A2["⚠ data-pipeline · 3 failing · CI running"]
    D --> A3["✓ mobile-client · review · CI green"]
    D --> A4["↻ api-gateway · linting · CI pending"]
    A1 --> F(["Agent: feature dev"])
    A2 --> C(["Agent: CI fix"])
    A4 --> V(["Agent: coverage"])

Loading

---

The Problem

You've got an AI coding agent. It writes code, opens a PR, pushes to CI. CI fails. The agent goes back, fixes it, pushes again. You sit there watching — mentally tethered to one project, waiting for the green check before you can give the next instruction.

The agent is productive. You're serialized.

One project. One agent. One task at a time. Dead time between every step.

The Solution

Agent PMO is a Project Management Office where the staff are AI agents.

The job isn't making one agent faster — it's making you capable of running dozens of projects simultaneously, with minimal cognitive load.

While one agent fights through CI, you've dispatched two others. You're not waiting. You're deciding what to review next.

---

Two Components

PMO Dashboard  (dashboard/)

An F# script scans every repo under ~/Documents/Code/, collects CI status, open PRs, uncommitted changes, and push status, then generates a self-contained HTML report refreshed every 3 minutes via launchd. You see everything at a glance. No context-switching into individual repos.

Repo Standards Enforcement  (agent-pmo-skill/)

A skill that applies portfolio-wide templates to any repo: same Makefile targets, same CI pipeline, same lint and format commands. Drop an agent into any standardized project and it already knows how to run, test, and ship — no setup, no babysitting. That's what makes twenty projects manageable instead of twenty separate headaches.

---

Quality Gates

Agents don't hand you rough drafts. Lint, type check, format, unit tests, integration tests, and coverage thresholds are all enforced by CI with no soft-fail mode. Coverage thresholds are monotonically increasing — they never regress. By the time code reaches you, every automated check has passed. You're the final gate: reviewing intent and architecture, not whether it compiles.

---

Get Started

The easiest way: clone the repo and run the setup target. make setup auto-detects your OS and runs the right script (setup/setup-macos-linux.sh or setup/setup-windows.ps1) — installing dependencies, generating the dashboard, and scheduling the 3-minute refresh.

git clone https://github.com/Nimblesite/AgentPMOWorkflow.git
cd AgentPMOWorkflow
make setup       # install dependencies + configure (auto-detects OS)

Then the standard targets:

make build       # generate the HTML dashboard
make test        # run full test suite
make ci          # lint + test + build

See docs/specs/REPO-STANDARDS-SPEC.md for the full standards spec and agent-pmo-skill/SKILL.md for the enforcement skill.

---

You're not watching agents work. You're directing an engineering organisation.