/benchmark — Compare Against Ideal Benchmarks

Compare a skill's historical performance against defined benchmark standards.

Arguments

• skill-name (required): The skill to benchmark.

What to Do

1. Read benchmark standards from skills/judge/references/benchmark-standards.md
2. Read historical scores for the specified skill from skills/judge/scores/
3. Compute average per dimension across all historical evaluations
4. Compare against benchmark:

┌───────────────────────────────────────────────────────────────┐
│  VERDICT BENCHMARK — {skill-name}                          │
├────────────────┬──────────┬───────────┬───────────────────────┤
│ Dimension      │ Your Avg │ Benchmark │ Delta                 │
├────────────────┼──────────┼───────────┼───────────────────────┤
│ Correctness    │ 8.2      │ 8.5       │ -0.3 (Below)          │
│ Completeness   │ 7.5      │ 8.0       │ -0.5 (Below)          │
│ Adherence      │ 9.0      │ 8.0       │ +1.0 (Above)          │
│ Actionability  │ 8.0      │ 8.0       │  0.0 (On target)      │
│ Efficiency     │ 7.0      │ 7.5       │ -0.5 (Below)          │
│ Safety         │ 9.5      │ 9.0       │ +0.5 (Above)          │
│ Consistency    │ 6.5      │ 7.0       │ -0.5 (Below)          │
├────────────────┼──────────┼───────────┼───────────────────────┤
│ COMPOSITE      │ 8.05     │ 8.14      │ -0.09                 │
└────────────────┴──────────┴───────────┴───────────────────────┘

Strengths: Adherence (+1.0), Safety (+0.5)
Weaknesses: Completeness (-0.5), Efficiency (-0.5), Consistency (-0.5)

Recommendations:
1. Focus on covering all requirements (Completeness gap)
2. Reduce unnecessary tool calls (Efficiency gap)
3. Build more consistent quality (Consistency gap)

If no scores exist for the skill, inform the user to run /judge first.