| name | workshop |
| description | Diverge on a problem or opportunity — generate many genuinely different solution shapes before committing to one. Use after /diagnose surfaces a problem or opportunity, or when you have a problem and want to see the full range of ways to solve it before scoping. |
| user-invocable | true |
| argument-hint | [problem or opportunity] |
MANDATORY PREPARATION
Invoke /product-thinking — it contains the Context Gathering Protocol and the AI Slop Test. Follow the protocol before proceeding.
Mindset
A workshop is divergence. Its one job: take a single problem or opportunity and expose the full range of shapes a solution could take — so the choice of shape is deliberate, not the first plausible idea.
The two failure modes:
- First-idea lock-in. The team names a problem, pictures one solution, and scopes it. The workshop's job is to prove there were nine other shapes worth seeing first.
- Ten variations of one idea. Producing "10 options" that all feel like the same product to a user is not divergence — it's padding. If two options would feel identical in the user's hands, they count as one.
This is not brainstorming. Brainstorming produces 40 ideas and zero decisions. A workshop produces genuinely different options and a recommended direction. Push back on weak options. Kill duplicates. The output is a clear map of the solution space plus a pointer to where to go next.
Input
A workshop needs one problem or opportunity to diverge on. It comes from:
/diagnose — a problem or opportunity it surfaced (the common path). Read the referenced report in .acumen/reports/..acumen/value-chain.md — an opportunity listed against an extension point in the value chain.- Directly stated — the user hands you a problem in the argument.
If the input is vague ("engagement is low"), sharpen it to a one-line problem statement before diverging — a fuzzy problem produces fuzzy options. If you cannot sharpen it from context, run /diagnose first and say so.
Context Pull
Load context as fuel for divergence, not to discover the problem (that was /diagnose's job):
- Product context. Read
.acumen.md — thesis, positioning, stage, constraints. Options that dilute the thesis still belong on the board; flag them, don't hide them.
- Personas. Read
.acumen/personas.md — whose problem this is, their workarounds, their daily workflow.
- Features. Read
.acumen/features.md — what exists today that an option could repurpose, extend, or replace.
- Value chain. Read
.acumen/value-chain.md — where in the workflow the problem sits and which adjacent steps an option could reach into.
- Competitors. Read
.acumen/competitors.md — how others solve this shape of problem, and where the white space is.
If a context file is missing, note it and diverge from what you have. Do not fabricate context.
The Divergence Engine
Generate 10 genuinely different solution options for the one problem. Difference is the whole point — steer with two axes and a set of generators, then prune duplicates.
Axis 1 — Scope (how big a move)
Span the ladder from smallest to boldest. Aim for coverage across bands, not a cluster in the middle. If you skip a band, say why.
- Band-aid — config, copy, default flip
- Tweak — repurpose an existing capability
- Feature — new capability inside an existing surface
- Surface — new view, page, or workflow
- Platform — changes the product's shape across personas/features
- Category — redefines what the product is
Axis 2 — Nature (what character it has)
Give each option a distinct character. Don't repeat a Nature without a reason.
UX · automation · AI · integration · content · pricing · ecosystem · service
Divergence generators
When options start converging, force new shapes with these prompts (borrowed from /diagnose's reframing lenses):
- Inversion — what if we did the opposite? Removed the step instead of improving it?
- Substitution — what if this were a service instead of a feature? A template? A default? A one-time setup instead of an ongoing tool?
- Audience shift — what if this were only for power users? Only for brand-new users? Built for the admin, not the end user?
- Constraint removal — unlimited eng time: what's the boldest shape? Zero eng time: what could ship this week? The two extremes are rarely the same option.
- Steal from adjacent — how would a product in a completely different category solve this shape of problem? Borrow their move.
Collapse test
Before finalizing, apply it to every pair: would these feel like the same product in the user's hands? If yes, merge them and generate a replacement from an untouched Scope band or Nature. Ten real options beats twelve with two duplicates.
For each option
- Shape — one sentence: what the user would see or do
- Scope — band-aid / tweak / feature / surface / platform / category
- Nature — pick one
- Why this shape — one line: what makes this orientation worth considering
- ASCII sketch — materialize the option as a rough drawing so the shape is visible, not just described. Sketch the key screen, flow, or interaction. The bolder the option (surface / platform / category), the more the sketch earns its place — prose alone hides what makes it different. Keep it loose: layout and the one distinct element, not pixel fidelity.
+------------------------------------------+
| [What the user sees for this option] |
| |
| [The one element that makes it distinct]|
+------------------------------------------+
Converge
Ten options is a map, not a decision. Narrow it:
- Rank on a 2x2 — impact on the problem (how much would this actually move it?) vs. thesis fit (how well does it align with where the product is going?). Top-right is where to focus.
- Recommend a direction — one option, or a sequence ("band-aid now to stop the bleeding, surface later to own the workflow"). Say why the shape wins, not just the idea.
- Name what you're deliberately ignoring — the options you're setting aside and why. This stops the same shapes from resurfacing without new evidence.
Handoff
Point to the next skill explicitly:
/increment — scope the chosen option into a shippable spec. This is the default next step./roadmap — if the recommendation is a sequence of options over time./diagnose — if you need to baseline the metric the chosen shape should move, so you'll know whether it worked.
Output Format
Workshop: [Problem or Opportunity]
Diverging on: One-line problem/opportunity statement.
Source: /diagnose report, value-chain opportunity, or stated directly.
Options Board
| # | Shape | Scope | Nature | Why this shape |
|---|
| 1 | | band-aid / tweak / feature / surface / platform / category | UX / automation / AI / integration / content / pricing / ecosystem / service | |
(10 rows. Coverage across the Scope ladder — note any band deliberately skipped.)
Option Sketches
For each option, the ASCII sketch beneath its number and shape:
Option [#] — [Shape] · [Scope] · [Nature]
+------------------------------------------+
| [What the user sees for this option] |
| |
| [The one element that makes it distinct]|
+------------------------------------------+
Convergence (2x2)
Place the options on impact × thesis-fit. Call out the top-right quadrant.
Recommendation
Which option (or sequence), and why the shape wins.
Next step: /increment on [chosen option], /roadmap, or /diagnose to baseline the metric.
Deliberately Ignored
Options considered and set aside, with reasoning. Prevents them resurfacing without new evidence.
Save workshop output to .acumen/reports/workshop-[topic]-[date].md.
If divergence revealed a new extension point or opportunity not in .acumen/value-chain.md, add it there so /diagnose and future workshops can pick it up.
