creo-perflisted

# creo-perf — performance optimization with proof No performance claim without a number. Baseline before touching code, an identical re-measure after, one concern per commit, correctness checked on every pair. The harness makes the numbers deterministic; this skill makes the workflow non-negotiable. ## Commands | Command | What it does | |---------|--------------| | `/creo perf init` | Scaffold `.claude/skills/creo-perf/` for this project, detect the stack, fill targets, propose scenarios from the route code, write SQL sources, verify with preflight. **Run first.** | | `/creo perf preflight` | Tools, targets, DB, scenarios, SQL mode — everything a capture needs | | `/creo perf baseline [label]` | Full sweep under one immutable label (default `before`): platform, schema, every scenario, fe, workload | | `/creo perf audit [baseline\|after\|discover]` | Spawn `creo-perf-audit` (measure-only) for a severity-ranked findings report over the six layers | | `/creo perf optimize [scenario or finding]` | The fix loop: pick a finding, implement one concern, re-measure, verify correctness, commit | | `/creo perf scenario <id> <label>` | Re-capture one scenario | | `/creo perf fe <label>` | Initial-load / build / Lighthouse capture | | `/creo perf platform <label>` / `schema <label>` | Config-fact captures (context, not before/after metrics) | | `/creo perf workload <label>` | Discovery window: pg_stat_statements ranking under scripted traffic | | `/creo perf after <label> [ids...]` |