Read the docs folder first.

Metadata

• Stable ID: codex-user-prompt:b5161ded533ee0f2
• Source kind: codex-session-user
• Category: coding-agent-workflow
• Timestamp: 2026-05-14T20:50:42.354Z
• Semantic hash: b5161ded533ee0f2b68df8969e54efe2ae22827b34090f781903e010f54e938c
• Public handling: selected full-text prompt with secret filtering.

Prompt Text

Read the docs folder first.
 
Task:
Implement the AI workflow layer.
 
Scope:
- packages/ai
- LLMProvider interface
- MockLLMProvider
- WorkflowRunner
- AgentRunService
- SourceReference service
- Zod schemas
- SafetyGuard interface
 
Non-goals:
- Do not connect real OpenAI yet.
- Do not build chat UI.
- Do not put model calls inside UI or API routes.
 
Validation:
A mock workflow run creates AgentRun, validates structured output, creates SourceReference, and stores safety status.

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.

Related

• codex-prompt-corpus
• codex-prompt-taxonomy