User Stories
When To Use
Engineers say the requirements are not detailed enough but not what is missing. Use it before refinement, on one PRD section or one story map slice. It answers: which stories can the team build and test, and which questions must be settled in refinement first?
When Not To Use
If there is no agreed problem or requirement yet, stories will only formalise a guess; start with PRD Template. If the question is which stories come first, lay out the journey with Story Mapping.
Inputs
- The requirements (PRD section) or the story map slice
- The team's Definition of Done, if it has one
- Known constraints: data, integrations, permissions, platforms
If you have none of this, I start from one requirement in a sentence and mark the output as a first draft.
Approach
Stories take the common form "As a [role], I want [capability], so that [outcome]". Each is checked against INVEST, from Bill Wake (https://xp123.com/invest-in-good-stories-and-smart-tasks/). Acceptance criteria follow Gherkin's Given / When / Then (https://cucumber.io/docs/gherkin/reference/), and the Definition of Done follows the Scrum Guide 2020 (https://scrumguides.org/scrum-guide.html). The failure it prevents: criteria that script the screen ("click the blue button") and still never say what should happen when the payment fails.
Workflow
- Ask three questions: which slice or requirements, which roles use this, and is there a Definition of Done?
- Slice each requirement into stories by user outcome, not by layer ("front end", "API" are tasks, not stories).
- Write each story as "As a [role], I want [capability], so that [outcome]". If the "so that" is empty, the story has no value yet.
- Acceptance criteria per story: Given (context), When (one event), Then (observable outcome), with And / But; three to five steps per scenario. Use a Scenario Outline with an Examples table for data variations. Describe behaviour, never clicks.
- INVEST check: Independent, Negotiable, Estimable, Valuable, Small (at most a few person-weeks, per Wake), Testable. Flag each failure with the reason and a split or fix.
- Refinement questions: everything the story cannot answer (error cases, empty states, permissions, limits). I flag these; I do not settle them. Link the Definition of Done so criteria do not repeat it.
Output Format
# User Story Set
**Source:** [PRD section or slice] | **Definition of Done:** [link or none]
## Story [ID]
As a [role], I want [capability], so that [outcome].
Scenario: [behaviour]
Given [context]
When [event]
Then [observable outcome]
## INVEST check
| Story | I | N | E | V | S | T | Flag and fix |
|---|---|---|---|---|---|---|---|
| [ID] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [split or fix] |
## Questions for refinement
| Story | Question | Who answers (role) |
|---|---|---|
| [ID] | [question] | [role] |
## Decision
[Named person] decides at refinement on [date] which stories are ready for planning.
Done When
- Every story has a role, a capability and a non-empty "so that"
- Every story has at least one scenario with an observable Then
- Every INVEST flag carries a proposed split or fix
- Open points sit in the refinement table, not buried in criteria
Quality Bar
- Criteria describe behaviour, not UI steps, with one event per When (two events means two scenarios)
- No invented data values, limits or thresholds: [placeholders] until engineering or the user confirms
- Estimates belong to the team; I never size stories in points or days
- Stories are refined with engineers; Claude flags gaps, it does not settle them
Next
Run pm-pre-mortem (Pre-Mortem Analysis) to check what could go wrong before the date is set.
About the makers
This pack is made by 90n (90n.ai): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
1---2name: pm-user-stories3description: Writes a User Story Set from requirements or a story map slice, with Given / When / Then acceptance criteria, an INVEST check per story and questions for refinement. Use for "run pm-user-stories", "write user stories", "acceptance criteria", "Given When Then", "INVEST check", "engineers say the requirements are not detailed enough", "prepare for backlog refinement", part of the AI for Product Management Pack by 90n.4metadata:5 source: https://skillmd.com/skills/polar-bear-org/pm-user-stories6---78# User Stories910## When To Use11Engineers say the requirements are not detailed enough but not what is missing. Use it before refinement, on one PRD section or one story map slice. It answers: which stories can the team build and test, and which questions must be settled in refinement first?1213## When Not To Use14If there is no agreed problem or requirement yet, stories will only formalise a guess; start with PRD Template. If the question is which stories come first, lay out the journey with Story Mapping.1516## Inputs17- The requirements (PRD section) or the story map slice18- The team's Definition of Done, if it has one19- Known constraints: data, integrations, permissions, platforms20If you have none of this, I start from one requirement in a sentence and mark the output as a first draft.2122## Approach23Stories take the common form "As a [role], I want [capability], so that [outcome]". Each is checked against INVEST, from Bill Wake (https://xp123.com/invest-in-good-stories-and-smart-tasks/). Acceptance criteria follow Gherkin's Given / When / Then (https://cucumber.io/docs/gherkin/reference/), and the Definition of Done follows the Scrum Guide 2020 (https://scrumguides.org/scrum-guide.html). The failure it prevents: criteria that script the screen ("click the blue button") and still never say what should happen when the payment fails.2425## Workflow261. Ask three questions: which slice or requirements, which roles use this, and is there a Definition of Done?272. Slice each requirement into stories by user outcome, not by layer ("front end", "API" are tasks, not stories).283. Write each story as "As a [role], I want [capability], so that [outcome]". If the "so that" is empty, the story has no value yet.294. Acceptance criteria per story: Given (context), When (one event), Then (observable outcome), with And / But; three to five steps per scenario. Use a Scenario Outline with an Examples table for data variations. Describe behaviour, never clicks.305. INVEST check: Independent, Negotiable, Estimable, Valuable, Small (at most a few person-weeks, per Wake), Testable. Flag each failure with the reason and a split or fix.316. Refinement questions: everything the story cannot answer (error cases, empty states, permissions, limits). I flag these; I do not settle them. Link the Definition of Done so criteria do not repeat it.3233## Output Format34```markdown35# User Story Set36**Source:** [PRD section or slice] | **Definition of Done:** [link or none]37## Story [ID]38As a [role], I want [capability], so that [outcome].39 Scenario: [behaviour]40 Given [context]41 When [event]42 Then [observable outcome]43## INVEST check44| Story | I | N | E | V | S | T | Flag and fix |45|---|---|---|---|---|---|---|---|46| [ID] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [ok / flag] | [split or fix] |47## Questions for refinement48| Story | Question | Who answers (role) |49|---|---|---|50| [ID] | [question] | [role] |51## Decision52[Named person] decides at refinement on [date] which stories are ready for planning.53```5455## Done When56- Every story has a role, a capability and a non-empty "so that"57- Every story has at least one scenario with an observable Then58- Every INVEST flag carries a proposed split or fix59- Open points sit in the refinement table, not buried in criteria6061## Quality Bar62- Criteria describe behaviour, not UI steps, with one event per When (two events means two scenarios)63- No invented data values, limits or thresholds: [placeholders] until engineering or the user confirms64- Estimates belong to the team; I never size stories in points or days65- Stories are refined with engineers; Claude flags gaps, it does not settle them6667## Next68Run pm-pre-mortem (Pre-Mortem Analysis) to check what could go wrong before the date is set.6970## About the makers7172This pack is made by 90n ([90n.ai](https://90n.ai)): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).