| name | slo-pattern |
| description | Use this skill after /slo-play has filled §4, to fill §5 Pattern Catalog — turn raw play into named reusable mechanisms (a vocabulary of tricks). Mode is convergent. Cite probe IDs for every pattern, run the next-curve check and the DICEE check, and narrow to at most five serious candidates. Do NOT promote everything. Hands off to /slo-precision.
|
/slo-pattern — name the reusable tricks
The play log is full of raw probes. Your job is to name the reusable mechanisms
hiding in them — the "vocabulary of tricks" — and narrow the field enough to
measure. Mode: convergent (judge reusability).
Inputs
docs/slo/experiments/<slug>/EXPERIMENT.md — §4 Play Log (probe ledger, dead-ends).- The binding spec
docs/slo/design/innovation-loop-experiment-book-spec.md (§5 body).
Output
Fill §5 Pattern Catalog of the Experiment Book. Produces PatternCatalog
(+ NextCurveCandidates, ProductPull, ArchitecturePull).
Method — fill §5, opening with the Phase Contract (Mode: convergent)
- Pattern candidates table: name, mechanism, probe evidence (cite probe
IDs), why surprising, reuse cases, risks. Cite probe IDs for every
pattern — a pattern with no probe evidence is rejected.
- Cap: at most five serious candidates (≤5). If play produced more, narrow.
- Next-Curve check: for each pattern, is it a 10% improvement or a
category change? (Kawasaki's "jump to the next curve".)
- DICEE check: Deep / Intelligent / Complete / Empowering / Elegant.
- Sunlit strategic fit: B2C / B2B / secure-data / cybersecurity.
- Product pull + Architecture pull: which patterns suggest a user-facing
wedge vs. a reusable platform capability.
- Update the §1 tracker (phase 3
complete), append a Safety Check, hand off.
Gate
Do not proceed with more than five serious candidates, and never with a pattern
that cites no probe ID. The output should narrow the field enough to measure.
Handoff
Suggest /slo-precision <slug> — make the surviving claims measurable.
Anti-patterns
- Promoting everything (the cap exists to force a choice).
- Naming a pattern with no probe evidence (cite probe IDs).
- Confusing a 10% improvement with a next-curve move — the next-curve check
exists to keep that honest.
Loops: Innovation Sandbox loop — see docs/LOOPS-ENGINEERING.md#innovation-sandbox-loop.
