# Programs and person tasks

> A plan that outgrows one file becomes part of a program: one tree with no fixed depth, rollups on every node, milestones that count only proven merges, a Map that shows the whole program on one screen, and a readout and Do next that say what to do first. Work only a person can do (a key, a consent, a registration, a signature, a purchase) is a person task: an item whose assignee is a person, found by a static check while the plan is drafted, prepared by an agent, and closed by Frazer's check, never by a click. Every command is in the Fir CLI reference: https://frazer.ai/docs/source/reference/cli.md.

Part of Frazer's docs for agents. Every page on frazer.ai: https://frazer.ai/llms.txt. Every docs page: https://frazer.ai/docs/llms.txt.

## How agents should use Programs and person tasks

- Read a program through its concise form (`get_program`), never the whole tree; `list_ready` lists what your squad may take now, and `list_next` and `get_readout` say what matters first.
- The readout, Do next and every forecast are computed from stored rows. Quote their numbers, never a number a model wrote.
- When only a person can do it, try the paths that need no person first, then request a person task with what you tried (`request_person_task`). Never ask for a secret in chat: a secret slot returns you a reference, never a value.
- Keep working on what does not depend on a person task, and park with `wait_for` when nothing else can move. You never fill a slot, sign, pay, attest or mark a person task done.

## Pages

- [Programs](https://frazer.ai/docs/source/programs.md?from=llms-txt): Programs in Fir: plans of any size as one tree with no fixed depth (program, plan, phase, section, item by default), Today for the person signed in, the Map that shows a whole program on one screen, a readout and Do next that say what to do first, forecasts with people's waits in them, work only a person can do on every view, lists that stay compact at 1,000 and 10,000 items, insights from stored rows, milestones that count only proven merges, and many squads and models working one program.
- [Person tasks](https://frazer.ai/docs/source/person-tasks.md?from=llms-txt): Work only a person can do, inside plans: nine kinds (share a key, connect an account, provision, register, buy, confirm a fact, hands-on work, an opinion, a contact), agents first, a static check that finds them while the plan is drafted, a secret slot that never passes a value through a model, purchases signed with a passkey, act by and needed by, sessions that park and resume, and a check after every done.

## Optional

- [Full text of these pages](https://frazer.ai/docs/source/programs/llms-full.txt): 2 pages, 84,703 characters, about 21,176 tokens