Book + templates
Running a Fleet of Coding Agents
One coding agent is a tool. Five in parallel are an operations problem — and most teams find out in week two, when two agents ship conflicting changes, a review bot silently skips, or a "healthy" chat starts confidently inventing APIs that don't exist.
This kit is the operating model that absorbs those failures. Nine short chapters cover the five ways fleets fail and the discipline that prevents each one: a single human merge authority, serialized critical paths, review protocols that survive bot limits, health thresholds that tell you when to kill a chat, and extra gates for code that moves money or mints accounts. It's distilled from running multi-agent workflows against production systems — no prompt packs, no hype.
Zip of markdown chapters and templates plus a compiled PDF. 7-day refund policy.
What you get
- A one-page merge checklist that turns "the checks look green" into an actual definition of done — print it, pin it, use it on every agent PR
- An always-on agent rule template that stops agents from merging, deploying, or treating a skipped bot as approval
- Health thresholds that catch context rot before the agent starts rewriting APIs that don't exist
- A handoff brief template so killing a bloated chat costs you nothing — the fresh chat picks up the work without re-litigating decisions
- The two test categories bots won't write (idempotency and concurrent-claim) as merge gates for anything that moves money or mints identity
- A three-layer knowledge system (rules / playbooks / handoffs) so your fleet stops re-learning the same lessons at your token expense
- MCP and credential hygiene rules that keep agents from ever seeing secret values or deploying to the wrong account
- A 90-minute adoption quickstart — rule installed, merge gate live, first handoff written, in one afternoon
Who it's for
- Tech leads running two or more coding agents (Cursor, Claude Code, Codex) against the same repos
- Solo founders who are their own merge authority and can't afford a double-mint incident
- Teams adopting agents who hit chaos in week two and want a system, not vibes
It's not for people looking for prompt packs, orgs that need a managed multi-tenant agent platform, or anyone still running a single chat at a time — bookmark it for when you scale.
FAQ
- What format is it?
- A zip of markdown files — nine chapters plus ready-to-copy templates — and a compiled PDF of the book. The templates are meant to be pasted into your repos and edited, so markdown is the primary format.
- Which tools does it apply to?
- The model is tool-agnostic: Cursor, Claude Code, Codex, or any setup where multiple agents produce PRs. Templates ship in Cursor's rule format with notes for adapting to other editors and to GitLab.
- Do I get updates?
- Yes — lifetime updates to this edition. Every revision is a free re-download from your Gumroad library.
- Is this just generic AI advice?
- No. It's distilled from operating multi-agent workflows against production systems, including the failure modes — merge stomps, silent bot skips, double-mints — that only show up when real code ships.
- What's the refund policy?
- 7 days, for buyers the kit genuinely didn't help. Requests after downloading the full bundle are reviewed individually, and refunded purchases lose access to future updates.