benchmark_template.md
2.04 KB
{Skill Name} — Benchmark v1.0.0
Last updated: {date} Benchmark version: 1.0.0 Trigger: initial_creation Previous version: — Change summary: Initial benchmark derived from knowhow.md and deep research.
Scoring Dimensions
| # | Dimension | Weight | What It Measures |
|---|---|---|---|
| 1 | {dim_1} | {%} | {description} |
| 2 | {dim_2} | {%} | {description} |
| 3 | {dim_3} | {%} | {description} |
| 4 | {dim_4} | {%} | {description} |
| 5 | {dim_5} | {%} | {description} |
| Total | 100% |
Detailed Rubric
Dimension 1: {name} ({weight}%)
| Score Range | Label | Criteria |
|---|---|---|
| 90–100 | Excellent | {specific observable criteria} |
| 70–89 | Good | {specific observable criteria} |
| 50–69 | Acceptable | {specific observable criteria} |
| 30–49 | Poor | {specific observable criteria} |
| 0–29 | Failing | {specific observable criteria} |
Dimension 2: {name} ({weight}%)
[repeat structure above]
Scoring Formula
Final Score = Σ (dimension_score × dimension_weight)
Example:
- Dim 1 score: 80 × 30% = 24
- Dim 2 score: 70 × 20% = 14
- ...
- Final = sum of all weighted scores
Pass Thresholds
| Threshold | Score | Meaning |
|---|---|---|
industry_avg_score |
TBD (set after baseline crawl) | Minimum to mark skill as enable |
elite_avg_score |
TBD (set after baseline crawl) | Target for excellent rating |
Scoring Notes
- Score each dimension independently before computing weighted sum
- When in doubt between two score bands, cite specific evidence from the output
- For each scored sample, record: total score, per-dimension breakdown, and 1-sentence justification per dimension