| name | genesis-ai-provider |
| description | Define reliable AI provider harnesses with mock providers, schemas, retries, invalid JSON recovery, persistence tests, and provider contracts. Use for LLM, image, voice, subtitle, render, and pipeline provider work. |
AI Provider Skill
Purpose
Make AI provider behavior deterministic under tests and recoverable under failures.
When to use
Use before adding or changing AI providers, provider adapters, retry logic, schema parsing, or persistence.
When NOT to use
Do not use for purely static documentation changes.
Inputs required
Provider name, request shape, expected response, error modes, retry policy, and persistence target.
Outputs required
Provider contract, mock provider fixture, schema validation, retry test, invalid JSON recovery test, and persistence test.
Required tests
Mock success, provider error, retry, invalid JSON recovery, and persistence tests.
Required fixtures
Create provider fixtures under fixtures/agents, fixtures/images, fixtures/tts, or related directories.
Required contract updates
Update contracts/agents/ or provider-specific contracts.
Required codebase map updates
Update providers summary, pipeline flow, and test matrix.
Token saving rules
Read provider summary and contract before provider code.
Acceptance criteria
Provider output is schema-validated and failure recovery is tested.
Common mistakes
Trusting raw model JSON, skipping retry tests, and using live provider calls in normal tests.
Recovery workflow
If provider output is invalid, capture fixture, add parser recovery test, then patch parser or contract.
