Rule: you write every line yourself. An AI may check spelling and format after you decide. It may not decide.
Epic
Date written: <YYYY-MM-DD> Author (human): <your name>
Title (the outcome, in six words):
Product:
Customer problem, in my own words (3 to 5 sentences):
Who is the user? Which journey step does this help?
Why now? What happens if we do nothing?
Repos I expect (count them):
- <repo> (kind) planned | exists
Definition of done (testable, three to five lines):
1.
2.
3.
Out of scope (so the agents do not guess):
-
Risks I accept (with my name):
- <risk> accepted by <name>Roadmap
| # | Epic | Owner | Start | Due | Depends on | Status | Why this place in the order |
|---|---|---|---|---|---|---|---|
| 1 | |||||||
| 2 |
Add a short list: What I chose not to build, and why.
Milestones for the first epic (3 to 5)
| # | Milestone | Repo | Due | Done when |
|---|---|---|---|---|
| 1 |
After you decide
You may now ask an AI agent to turn your milestones into issues. The agent builds. You review every merge request. You accept or reject.
Example (TrailMug, fictional)
Title: Visitors can request a welcome cup
Problem: People like the idea of a reusable cup but have no easy first
step. They leave the site without a clear next action.
Who: Maya, at the "find" step.
Repos: trailmug-site (exists), trailmug-svc-leads (planned),
trailmug-crm (planned)
Done:
1. A visitor submits an email on the page
2. The lead is in the customer list within 1 minute
3. A welcome message arrives within 5 minutes
Out of scope: payments, sign-in
Risk: the email provider may delay messages. Accepted by Maya.
