Anthropic provider fails on all semantic analyzers: endpoint rejects JSON Schema minimum/maximum keywords #66

With SKILLSPECTOR_PROVIDER=anthropic, every LLM-backed analyzer fails with a 400 and falls back, so semantic analysis silently degrades to static-only.

export SKILLSPECTOR_PROVIDER=anthropic
export ANTHROPIC_API_KEY=sk-ant-**
skillspector scan ./tests/fixtures/safe_skill --verbose

WARNING [skillspector.nodes.analyzers.semantic_quality_policy] semantic_quality_policy failed:
Error code: 400 - {'error': {'code': 'invalid_request_error',
'message': "For 'integer' type, property 'minimum' is not supported", ...}}

After fixing the integer field, the same error recurs for 'number' ("properties maximum, minimum are not supported"), and again separately in nodes.meta_analyzer (LLM call failed, using fallback).

Cause

The structured-output / tool schemas are generated from Pydantic models using Field(ge=..., le=...). Pydantic serializes these to JSON Schema minimum/maximum, which the Anthropic provider's endpoint rejects on integer/number types, 400-ing the whole request.
Affected fields:

src/skillspector/llm_analyzer_base.py:64 — start_line: int = Field(ge=1, ...)
src/skillspector/llm_analyzer_base.py:66 — confidence: float = Field(ge=0.0, le=1.0, ...)
src/skillspector/nodes/meta_analyzer.py:66 — confidence: float = Field(ge=0.0, le=1.0, ...)