Orchestration
Big work becomes a map of tickets that a conductor thread works through, and side work gets its own thread.

How it works
- A large change
- Map of tickets
- Conductor thread
- One worker thread per ticket
- Draft pull requests
Workflows
Planning maps
On with the packBig work becomes a map of tickets, and a conductor thread works through it with one worker thread per ticket.
Side threads
On with the packWhen you agree to a side fix, it starts in its own T3 thread, and this thread stays on its task.
Question guard
On with the packA top-level T3 thread that ends its turn with a decision for you in prose is stopped once and told to ask with the question tool, so the decision reaches you even while the thread waits on a watch.
Quality loop
Off until you turn it onNeeds an always-on devboxEach night, agents improve one area of a repo through small, proven draft pull requests that you review.
Read before you turn it on
Spends your AI plan's usage every night.
Signal loop
Off until you turn it onNeeds an always-on devboxEach night, agents turn production errors and friction that nobody handles into small draft fix pull requests.
Read before you turn it on
Reads production signals and spends your AI plan's usage every night.
PR digest
Off until you turn it onNeeds an always-on devboxEach weekday morning, one thread lists your open pull requests and what blocks each one.
What it costs
The loops run on an always-on devbox and spend your AI plan's usage each night, so they start off.
Know before you turn it on
The loops open pull requests without you in the room. Each one stays a draft until you say otherwise.
Settings10 settings that your agent asks about
orchestration.trackerWhich repo's issues hold your maps? Leave it empty to use your context repo, or your Fleet profile repo when you have none.Default: from your context repoorchestration.lessonsLogWhich file holds the lessons log that maps read and write, such as a path in a private repo?Default: from your context repoorchestration.rolloutLogWhich file logs your rollouts, one JSON line per stage, so fleet lessons can name holds and rollbacks? ~/ is allowed.Default: from your context repoorchestration.goalsWhere do your team's goals live, so a map can name the metric its work moves?Default: from your context repoorchestration.analyticsProjectWhich analytics project shows how people use your product?orchestration.loopProjectWhich T3 project do the loops and the PR digest run in?orchestration.qualityAreaWhich paths of the repo may the loops change, such as src/billing?orchestration.criticalPathsWhich paths need your review before a loop's pull request is ready, such as src/payments?orchestration.ledgerRepoWhich private repo holds the loops' ledger? It records production numbers, so keep it private.orchestration.signalSourceWhere should the signal loop read production errors and friction? Leave it empty to use an analytics MCP when one is connected.
Add Orchestration to your Fleet
Paste this prompt into a T3 Code thread on a machine that runs Fleet.
Your install prompt
Add these Fleet packs to my setup with me: Pull requests and Orchestration. Fleet already runs on this machine. Ask me each choice with your question tool, and run every command yourself. 1. Run `fleet update`, so this machine has the newest engine and packs. Then run `fleet packs`, and leave out each pack that is already on. 2. Read the setup note of each pack. Ask me what it asks, and collect the flags that it names. Where a note names "the setup command" or "the command in step 9", use the `fleet packs` command of this prompt. Do not run a step that a note puts after that command yet. - Orchestration: ~/.local/share/fleet/engine/packs/orchestration/SETUP.md 3. Turn the packs on with one command, and add the flags that the notes name: fleet packs --add pull-requests --add orchestration 4. Do the steps that each note puts after the command. 5. Run `fleet doctor`, and explain each line. Do each `To do:` line with me, and say what a sign-in is for before it opens. 6. My other machines take the change at their next `fleet update`. To give it to one now, run `fleet update` there.