Read the docs folder first.

Metadata

  • Stable ID: codex-user-prompt:917ddae462b16361
  • Source kind: codex-session-user
  • Category: coding-agent-workflow
  • Timestamp: 2026-05-14T21:28:59.351Z
  • Semantic hash: 917ddae462b16361cf96428ea0c8fc709a9aea03f52a7c5af84ef67976db1028
  • Public handling: selected full-text prompt with secret filtering.

Prompt Text

Read the docs folder first.
 
Task:
Implement OpenAIProvider behind the existing LLMProvider interface.
 
Rules:
- No direct OpenAI calls from UI.
- No direct OpenAI calls from API routes.
- All calls go through WorkflowRunner and LLMProvider.
- Use structured output matching existing Zod schemas.
- Persist provider, model, prompt version, input hash, output, safety status, and errors in AgentRun.
- Keep MockLLMProvider available for local development.
 
Validation:
Existing mock workflow tests still pass.
A real provider can be enabled through env vars without changing workflow code.

Reuse Notes

  • EXTRACTED: This is a selected Codex prompt or automation prompt from the local Codex corpus.
  • INFERRED: Future agents can reuse its structure, constraints, and acceptance criteria when creating similar Codex workflows.