Claude Code orchestration, run like a team
One commander session plans the work, splits it into bounded task
packets, and judges the evidence. A pool of workers — subagents or
external CLIs like codex exec —
runs the implementation in parallel, each writing its own logs and
exit codes. Nothing merges until the commander has reviewed what
came back.
Paid products · Delivered instantly on Gumroad
Ebooks, prompt packs, and templates for the commander/worker way of running AI agents. Every purchase is covered by a 30-day refund window — if a product doesn't work for you, ask for your money back.
Bundle · best value
Everything on this page in one download: both books (EN + KO), the 25-prompt review pack, the 30-pattern CLAUDE.md library, and all templates in both languages — 82 files, with a START_HERE guide that maps each problem to the file that solves it. Buying the four separately costs $176.
Ebook
A practical guide to running Claude Code as a commander with a pool of parallel workers: splitting work, writing bounded task packets, and reviewing agent output before anything merges.
Prompt pack
Ready-to-use prompts for adversarial review of AI-written code, aimed at what agents actually ship: tests that pass without proving anything, swallowed errors, and quiet regressions just outside the diff.
Pattern library
Copy-and-adapt CLAUDE.md patterns: how to structure rules that agents actually follow, where verification commands belong, and what to leave out so the file stays short enough to be read.
Templates
Starter files for a commander/worker setup, so you can adapt a working structure to your own repository instead of starting from an empty file.
Free · Open repositories
The same patterns, free to read and reuse on GitHub — before you spend anything, check the quality against the real thing.
Free · on GitHub
Free · on GitHub
Adversarial review prompts you can adapt into checks and evidence clauses.
View on GitHubFree · on GitHub
Free rule patterns that demand evidence; the paid library expands the set.
View on GitHubFree · on GitHub
The free orchestration repository the articles and the book build on. Written in Korean.
View on GitHubFree guides
Long-form English guides on orchestrating parallel workers, reviewing AI-written code, and detecting stalled agents.
Codex operations
Preserve logs, PIDs, and exit codes per worker, and tell DONE, FAILED, and STALL apart with a watcher script.
Adversarial review
Exit codes, tests that never fail, migration lock-in, prompt injection, and doc drift — checked against real code examples.
Claude Code usage
The commander/worker split, native subagents versus external CLI workers, and measuring usage and rework together.
CLAUDE.md rules
The trigger, check, stop, evidence structure, five worked examples, and a probe for verifying rules in a fresh session.
Agent operations
Per-worker logs, PIDs, exit markers, and a portable watcher to tell done, failed, running, and stuck apart.
Not sure yet?
Read one article or browse a free repository first. The paid products cover the same ground in more depth, and every Gumroad purchase has a 30-day refund window.