Step 1 / 5 · 15 min
Bound one valuable behaviour before describing implementation.Cut the slice
- Write one user and one job.
- Choose a start trigger and observable end state.
- If the slice contains “and then”, test whether it should be split.
Shape · Core skill lab
One-Slice Executable Spec
Define one end-to-end behaviour tightly enough that a human or agent can build and verify it without guessing.
- Decision
- Is this workflow slice bounded, buildable, and testable enough to enter implementation?
- Timebox
- 75 minutes + one builder handoff
- Output
- Executable workflow spec
Bring to the bench
- A validated product bet
- A single user workflow to slice
- A reviewer who could implement it
Your work saves in this browser.
Field tools
Use the instrument, not a blank page.
Copy these into your interview, agent, review, or working document. They are specific to this repetition.
template
Behavioural acceptance pattern
Use one block per meaningful state transition.
SCENARIO: [observable outcome] GIVEN [starting state and permissions] AND [required data or dependency] WHEN [one user or system action] THEN [observable state] AND [quality, audit, or notification effect] BUT NOT [forbidden side effect]
prompt
Plan-first handoff prompt
Give this to the builder or coding agent before implementation.
Read this spec as an implementer. Do not write code. Return: (1) state model, (2) smallest implementation sequence, (3) ambiguous requirements, (4) missing failure modes, (5) acceptance criteria that cannot yet be verified, and (6) files/systems likely affected. Prefer questions over assumptions.
Calibrate judgment
Compare the evidence, not the polish.
A slice with edges
An AI-assisted handoff brief is being scoped.
Start when a signed contract exists; end when a delivery lead accepts a source-linked brief. If source documents are missing, the system asks for named inputs and preserves the draft. It never fabricates contract terms. No e-signature or template management in this slice.
Why it works: The start, end, trust failure, recovery, and non-goals can all be implemented and tested.
A feature-shaped spec
Build an AI onboarding experience that is intuitive, fast, delightful, and scalable. Add a dashboard and integrations where useful.
Why it fails: There is no actor, state model, observable acceptance, failure behaviour, or scope boundary.
Review → revise → repeat
The artifact is the beginning of the rep.
Check only standards your current artifact actually meets. Then record one consequential revision before exporting it.