Add Iris flowers dataset analysis tasks

[ScuttleBot]

Adds 3 new data-analysis tasks using the classic Iris flowers CSV dataset (assets/csvs/iris_flowers.csv):

• task_csv_iris_summary — Statistical summary: overall and per-species descriptive statistics, correlation analysis, and key findings
• task_csv_iris_classify — Species classification: develop threshold-based rules to classify species from measurements, evaluate accuracy with confusion matrix
• task_csv_iris_outliers — Outlier detection: IQR-based outlier analysis at overall and within-species levels, identify unusual observations

All tasks use category: data-analysis, grading_type: hybrid (60% automated / 40% LLM judge), and 180s timeout.

Closes #234, Closes #237, Closes #238

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

This is a well-structured addition of three Iris dataset benchmark tasks. The grading logic is sound, the automated checks cover the key expected values accurately, and the LLM judge rubrics are clearly defined and weighted appropriately. The source: csvs/iris_flowers.csv path convention is consistent with existing CSV tasks in the repo.

Files Reviewed (4 files)

• tasks/manifest.yaml
• tasks/task_csv_iris_summary.md
• tasks/task_csv_iris_classify.md
• tasks/task_csv_iris_outliers.md

---

Fix these issues in Kilo Cloud

---

_{Reviewed by claude-4.6-sonnet-20260217 · 166,831 tokens}

[ScuttleBot]

🧪 Test Started

Instance: 155.138.225.173 (Vultr vc2-2c-4gb, ATL)
Branch: tasks/csv-iris
Tasks: task_csv_iris_summary, task_csv_iris_classify, task_csv_iris_outliers

Models (running in parallel):

• openrouter/anthropic/claude-opus-4.6
• openrouter/openai/gpt-5.4
• openrouter/google/gemini-3.1-pro-preview

ETA: ~20-30 minutes

Started at: 2026-04-15 13:30 UTC
Triggered by: ScuttleBot 🦀