Ingesting Spec
Developmentingesting-specIngesting Spec is an Agent Skill for development published in the openai/openai-dotnet repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Guide for ingesting the latest OpenAI TypeSpec specification into the openai-dotnet SDK. Use this when asked to update or ingest OpenAI API specs, copy base TypeSpec files from upstream, fix client TSP compile errors, or run code generation for new API areas.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/openai/openai-dotnet/blob/HEAD/.github/skills/ingesting-spec/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/ingesting-spec, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Spec Ingestion
Overview
This skill describes how to ingest the latest OpenAI TypeSpec specification (from the upstream microsoft/openai-openapi-pr repository) into the openai-dotnet SDK, area by area.
The process involves:
- Copying updated base specs from upstream (exact copy, no modifications)
- Reporting any compile errors in the base TSP (do NOT fix — base spec must stay unmodified)
- Fixing compile errors in the client TSP layer
- Preserving custom C# code (renames, stubs)
- Running code generation
- Verifying the output
Skill Documents
This skill is split across multiple files for easier navigation:
| Document | Description |
|---|---|
| steps.md | Step-by-step process — the full 9-step workflow from copying spec to post-generation verification |
| file-locations.md | Key file locations — quick reference for all upstream and local paths, area mappings |
| patterns-and-gotchas.md | Common patterns & gotchas — lessons learned, pitfalls, and conventions to follow |
| checklist.md | Checklist — a task-by-task checklist for tracking progress during an ingestion |
| references.md | Reference PRs — detailed notes on past ingestion PRs with lessons learned |
Quick Start
- Review references.md for examples of past ingestions in your area
- Read file-locations.md to understand the repo layout
- Follow steps.md for the full ingestion workflow
- Use checklist.md to track your progress
- Consult patterns-and-gotchas.md when you hit issues
Available Areas
Areas that can be ingested independently:
administration · assistants · audio · batch · chat · containers · conversations · embeddings · evals · files · fine-tuning · graders · images · models · moderations · realtime · responses · runs · threads · vector-stores · videos
Key Rules
- Always add
@@clientLocationfor every operation in the client TSP (the latest spec no longer usesinterfaceblocks) - NEVER modify the base spec — it must be an exact copy of upstream. Handle all issues (type unions, suppressions, etc.) in
specification/client/instead - Update
[CodeGenType]stubs insrc/Custom/{Area}/Internal/GeneratorStubs.csfor any renamed types - Defer complex features — suggest them as follow-up items rather than implementing in the same ingestion
- Run
./scripts/Invoke-CodeGen.ps1to generate code (warnings are OK. Errors are not), thendotnet buildto verify (Should be no warnings or errors) - Work locally only — do NOT create PRs or file issues. Instead, suggest a list of issues that may need to be filed upstream