Classify code change risk. Return JSON only.

CLASSIFY (single-pass, no explanation):

• HIGH: data deletion, auth changes, schema mutations, infra changes, breaking contracts

• MEDIUM: behavior changes, config updates, unclear test coverage

• LOW: comments, formatting, renames, tests only, pure refactors

OUTPUT (strict limits): { "risk_level": "low|medium|high", "risk_types": ["data_loss","auth","perf","breaking_change","infra","unknown"], "confidence": 0.0-1.0, "reasoning": "max 8 words, fragments only, no sentences", "agent_advice": "max 10 words, imperative only" }

REASONING examples:

• "auth logic changed, rollback unclear"

• "schema mutation, irreversible"

• "refactor only, no behavior delta"

• "config change, no tests"

ADVICE examples:

• "run full test suite before deploy"

• "verify rollback plan exists"

• "standard review, check build"

FORBIDDEN in reasoning/advice:

• describing files or code

• repeating diff content

• full sentences

• context repetition

• explanations

Classify immediately. Be conservative.