OpenAI Documentation Access Skill

v2.88 Key Changes (MODEL-AGNOSTIC)

• Model-agnostic: Uses model configured in ~/.claude/settings.json or CLI/env vars
• No flags required: Works with the configured default model
• Flexible: Works with GLM-5, Claude, Minimax, or any configured model
• Settings-driven: Model selection via ANTHROPIC_DEFAULT_*_MODEL env vars

ultrathink - Take a deep breath. We're not here to write code. We're here to make a dent in the universe.

The Vision

Documentation lookups should feel inevitable and authoritative.

Your Work, Step by Step

1. Identify target: Pick the right library and query.
2. Resolve ID: Use Context7 to get exact docs.
3. Query precisely: Ask for focused, answerable snippets.
4. Apply results: Convert docs into actionable guidance.

Ultrathink Principles in Practice

• Think Different: Seek the most direct source of truth.
• Obsess Over Details: Use exact IDs and query wording.
• Plan Like Da Vinci: Decide the lookup path before querying.
• Craft, Don't Code: Keep results concise and relevant.
• Iterate Relentlessly: Re-query when ambiguity remains.
• Simplify Ruthlessly: Avoid unnecessary sources.

OpenAI Documentation Access Skill

Access OpenAI's complete developer documentation through Context7 MCP integration.

Quick Start

# Query Codex CLI documentation
mcp__context7__query-docs:
  libraryId: "/websites/developers_openai_codex"
  query: "your question about Codex CLI"

# Query OpenAI API documentation
mcp__context7__query-docs:
  libraryId: "/websites/platform_openai"
  query: "your question about OpenAI API"

Available Documentation Libraries

Common Queries

Codex CLI Configuration

# MCP server configuration
query-docs:
  libraryId: "/websites/developers_openai_codex"
  query: "MCP server configuration config.toml remote URL authentication"

# Sandbox modes and permissions
query-docs:
  libraryId: "/websites/developers_openai_codex"
  query: "sandbox modes read-only workspace-write danger-full-access"

# Session resume and parallel execution
query-docs:
  libraryId: "/websites/developers_openai_codex"
  query: "exec resume session parallel execution workflow"

OpenAI API Integration

# Chat completions
query-docs:
  libraryId: "/websites/platform_openai"
  query: "chat completions API streaming function calling"

# Embeddings and vector search
query-docs:
  libraryId: "/websites/platform_openai"
  query: "embeddings API text-embedding-3 vector dimensions"

# Assistants API
query-docs:
  libraryId: "/websites/platform_openai"
  query: "assistants API threads messages runs tools"

SDK Usage

# Python SDK async patterns
query-docs:
  libraryId: "/openai/openai-python"
  query: "async client streaming completions error handling"

# Node.js SDK TypeScript
query-docs:
  libraryId: "/openai/openai-node"
  query: "TypeScript types streaming response handling"

MCP Configuration Reference

Context7 (Already Configured)

# ~/.codex/config.toml
[mcp_servers.context7]
command = "npx"
args = ["-y", "@upstash/context7-mcp@latest"]

Remote MCP Servers

# Streamable HTTP server format
[mcp_servers.example]
url = "https://example.com/mcp"
bearer_token_env_var = "EXAMPLE_TOKEN"
http_headers = { "X-Custom-Header" = "value" }

STDIO Servers

# Local command-based server
[mcp_servers.local]
command = "npx"
args = ["-y", "@package/mcp-server"]

[mcp_servers.local.env]
API_KEY = "your-key"

Integration with Codex CLI Skill

This skill complements /codex-cli by providing documentation lookup:

1. Before execution: Query docs for correct flags/options
2. On error: Look up error codes and solutions
3. Configuration: Find correct config.toml syntax

Workflow Example

# 1. Query documentation for correct syntax
# Use mcp__context7__query-docs with libraryId="/websites/developers_openai_codex"

# 2. Execute Codex with correct options
codex exec -m gpt-5.2-codex --full-auto "task"

# 3. On error, query docs for solution
# Query: "error code XYZ troubleshooting"

Claude-Codex Documentation Bridge

When Claude needs OpenAI documentation:

1. Use Context7 MCP (already available in both Claude and Codex)
2. Query appropriate library based on topic
3. Apply findings to implementation

This eliminates the need for a separate "openaiDevelopers" MCP - Context7 provides all OpenAI documentation with 614+ Codex CLI snippets and 9,400+ API snippets.

See Also

• /codex-cli - Codex CLI orchestration skill
• /library-docs - General library documentation via Context7
• ~/.codex/config.toml - MCP server configuration