| name | slo-play |
| description | Use this skill after /slo-sandbox has filled §3, to fill §4 Play Log — generate many raw probes from the sandbox's probe seeds WITHOUT premature judgment. Mode is divergent: the goal is to map possibilities and surface surprises and dead ends, not to pick a winner. This is the joy-preserving phase. Do NOT rank, optimize, or turn a probe into a product plan — that is /slo-pattern's and /slo-curate's job. Hands off to /slo-pattern.
|
/slo-play — generate raw probes, defer judgment
You are playing, not planning. The sandbox is set; your job is to run lots of
small probes and write down what happened — especially the surprises and the
dead ends — without deciding which is "best" yet. Mode: divergent.
The one rule (the joy guard)
During this phase you judge safety only; defer quality judgment. No ranking,
no winner-picking, no "this probe is the one", no turning a probe into a product
plan. A dead end is a valid, valuable output. The ONLY judgment allowed here is
the safety check. (Convergence is /slo-pattern and /slo-curate's job — not
yours.)
Inputs
docs/slo/experiments/<slug>/EXPERIMENT.md — §3 Sandbox Charter + probe seeds.
- The binding spec
docs/slo/design/innovation-loop-experiment-book-spec.md (§4 body).
Output
Fill §4 Play Log of docs/slo/experiments/<slug>/EXPERIMENT.md. Produces the
ProbeLedger (probe cards) + DeadEndList + StrangeButInterestingList.
Method — fill §4, opening with the Phase Contract (Mode: divergent)
- For each probe seed (and new ones that emerge), run a probe and write a row in
the Probe Board: id, probe, type, setup, observation, surprise, reusable?,
safety note. Generate MORE probes than you think you need.
- Tag each probe with a frozen type:
mechanism_probe | interaction_probe | failure_probe | security_probe | data_probe | latency_probe | magic_probe | composition_probe (composition = combine two unrelated tricks).
- Raw observations: bullets, unfiltered.
- Strange but interesting: things not useful yet that might matter later.
- Dead ends: what failed, what it taught, any reusable fragment. Capture
these deliberately — a good play session produces dead ends.
- Candidate patterns: a raw list for
/slo-pattern (names only — do NOT
evaluate or rank them here).
- Update the §1 tracker (phase 2
complete), append a Safety Check, hand off.
Gate
The output is not "we found the good idea". The output is "we now have enough
material — surprises, dead ends, candidate patterns — to name patterns". If you
catch yourself ranking, stop: that is the next phase.
Handoff
Suggest /slo-pattern <slug> — name the reusable tricks (convergent).
Anti-patterns
- Behaving like
/slo-plan — optimizing, reducing, criticizing quality too early.
- Discarding a dead end (it is evidence, not waste).
- Collapsing a probe into a feature plan.
- Stopping after three probes — diverge widely; the vocabulary comes from volume.
Loops: Innovation Sandbox loop — see docs/LOOPS-ENGINEERING.md#innovation-sandbox-loop.