sotalisted

# SOTA — System Assessment + Self-Heal **Arguments received:** $ARGUMENTS Repeatable scorecard that rates your Claude Code setup against state-of-the-art benchmarks. **Self-heal runs by default** — any category below 9/10 gets fixed automatically until the floor is met. ## Flags | Flag | Description | | -------------- | -------------------------------------------------------------- | | `--score-only` | Score only — skip self-heal, just show the scorecard | | `--gaps` | Show detailed improvement suggestions per category (then heal) | | `--history` | Show score trend over time from sota-history.json | | (default) | Score + self-heal all categories below 9/10 + commit fixes | ## Target Floor **Every category must reach 9/10.** After scoring, implement fixes for all categories below 9. Re-score after fixes. Repeat until all categories are ≥9 or no further auto-fixes are possible (propose blockers to user). ## Assessment Process ### Step 0: Rubric Refresh + Fetch Latest CC Features **Rubric version: 3.0 | Last reviewed: 2026-07-15 | Baseline: Claude Code 2.1.210** > **Why 3.0 is a rewrite, not a bump.** Rubric 2.1 claimed to be fresh (its own > staleness check only fires at >30 days) while scoring people against a Q1 2026 > world. Its 13 categories contained **zero** of: agent teams, automatic memory, > the Workflow tool, auto mode, `/loop`, cron/Routines, plugin packagi