The harness-engineering mindset — Agent equals Model plus Harness — step 7 of 8

Write audit_harness(setup) that takes a dict describing a harness and returns a dict with two fields:

• score: int 0-100, higher means more complete
• gaps: list of strings, the missing pieces by canonical name

The six canonical pieces (in this exact order):

["config", "tools", "infra", "orchestration", "hooks", "observability"]

Scoring rules:

• Each piece present adds 16 to the score (six pieces × 16 = 96).
• If ALL six pieces are present, return score 100 (rounding bonus).
• Missing pieces go into gaps in the canonical order above.

Two harnesses run for you. Expected output:

Claude Code: {'score': 80, 'gaps': ['observability']}
Raw API:     {'score': 0, 'gaps': ['config', 'tools', 'infra', 'orchestration', 'hooks', 'observability']}