Ticket template for AI agents: no room for the reasonable guess
A copyable ticket template for AI coding agents: validated frontmatter, a measurable acceptance criterion, an explicit out of scope list and the decisions already taken, with a filled example.
How to use it
- The human writes the ticket. The lead agent decomposes it. An agent writing its own acceptance criterion is grading its own homework.
- Every line under "What done looks like" has a number, a role or a test in it. If it does not, it is not a criterion.
- Open questions means the ticket is blocked, not started.
Blank: ticket.mdDownload
---
id: <number>
title: <verb + object, one line>
status: todo
area: <area>
depends_on: [<ticket ids>]
decisions: [<decision ids>]
board: <url of the row on the board>
---
## Why
One sentence linking this to the thesis.
## What done looks like (measurable)
- <role> does <action>; <observable result with a number or a test>.
- <the permission case: who must NOT see or do this, proven by a test that runs as them>.
- <the audit case: what is recorded, by whom, when>.
- Browser test on mobile and desktop passes.
## Out of scope (explicit)
- <adjacent feature> (ticket <id>).
- <adjacent feature>.
## Decisions already taken
- <id>: <one line>.
## Open questions
- None. (If there were, the ticket would be blocked, not started.)Filled example: ticket-example.mdDownload
---
id: 143
title: Invite a teammate to a collection
status: todo
area: team
depends_on: [131]
decisions: [0088, 0091]
board: <url of the row on the board>
---
## Why
Shared collections are the thesis for teams: one record, several people, every change attributed.
## What done looks like (measurable)
- A member with the owner role sends an invitation to an email; the
invitee receives it within 60 seconds in the test mailbox.
- A pending invitation grants zero rows of any table, proven by a test
that runs as the pending user.
- Accepting makes the invitee a member with the role chosen by the
owner; the audit log records who invited and who accepted, with
timestamps.
- Browser test on mobile and desktop passes.
## Out of scope (explicit)
- Resending or expiring invitations (ticket 144).
- Inviting by link.
- Any change to the roles themselves (decision 0088 fixes them).
## Decisions already taken
- 0088: the four roles and what each may do.
- 0091: invitations are per collection, not per account.
## Open questions
- None.Prompt to use it
Paste before the agent starts a ticket
Before writing any code, review ticket <id> against CLAUDE.md and the decisions log.
Return the ticket to me, without starting, if any of these is true:
- A line under "What done looks like" has no number, role or test in it.
- Two lines contradict each other, anywhere in the ticket.
- The work would touch something listed under "Out of scope".
- It needs a product decision that is not in "Decisions already taken".
If you return it, use the escalation format: one DECISION per gap, three options, your recommendation.
If none is true, reply READY and list the tests you will write first.Why it looks like this, and where each part came from: Writing tickets for AI agents: the template.
Changelog
- v1.0Nov 5, 2026First public version. The template behind every ticket of a seven week build; the out of scope list was the single most effective scope control.
More from the kit
Template / Build
CLAUDE.md template
The seven sections every agent session inherits, in the order an agent reads them, with nine neutral hard rules to start from.
v1.0 / CLAUDE.mdFormat / BuildEscalation and decision format
How an agent stops at a real decision: an id, three options, a recommendation. You answer with a letter.
v1.0 / escalation.mdTemplate / LeadMVP scope one pager
Thesis, one workflow, invariants and what is not built. The page that stops a fast build from sprawling.
v1.0 / mvp-scope.mdWorking on something like this?
This is the method behind how I build: AI native development with human gates. Discuss a product idea → Or email me.