FIELD NOTES + 7 TEMPLATES · PDF
The Agent Fleet Playbook
How one person runs dozens of Claude Code agents from a phone, on one Mac mini, without them stepping on each other or burning the whole rate limit in minutes.
Get it · US$12
One-time payment through Stripe. Instant download. 7-day refund, no questions: reply to your receipt.
Why it exists
One agent is easy. Thirty agents sharing a repo and a notes folder is a different problem: double builds, reverted work, tests fighting for the CPU, and a usage window that disappears.
One "usage reset, continue" message woke about forty agents at once. Seven minutes later, 83% of the window was gone.
Every rule in the playbook is there because something like that happened first.
What is inside
- Lanes. The one rule that stops agents "helping" each other into a mess, and how to make it hold.
- The shared brain. Six markdown files every agent writes to and almost never reads. Write always, read by grep.
- Briefs and reports. A six-part brief, three end states, and the "what I did not verify" line.
- Builders, one lander, one gate. Worktrees, batching, bisecting, a pre-push hook, a heavy-run lock.
- Usage as the budget. Pacing against two windows, the 85% rule, waking in waves, cheap readers, compaction.
- Time. Why "I'll check back later" does nothing, and what to use instead.
- Safety rails worth having from day one, and a first-week plan.
The templates
CLAUDE.md HANDOFF.md LESSONS.md OWNER.md brief-template.md landing-protocol.md pacing.md
Plain markdown. Copy, change, use in your own projects.
What it is not
- Not a framework or a tool. There is no code to install.
- Not long: 7 pages plus the templates. It is the rules, not the story.
- Not affiliated with or endorsed by Anthropic.
Disclosure: the playbook was drafted by one of the agents in the setup it describes, from its own standing rules and logs. Limits and tools differ between plans; treat the numbers as examples.
Get the playbook · US$12