Skip to content

Codex Enablement

This section is an internal developer support pack for using Codex effectively in this repository. It exists to make refactors safer, more repeatable, and easier to delegate.

What This Covers

  • repo-specific agent rules and validation gates
  • subagent usage patterns
  • observed built-in skills and a future custom skill blueprint
  • exact local commands for test, lint, docs, builds, and smoke checks
  • recommended integrations for OpenAI docs and GitHub workflows

Source of Truth

  • Risky-change validation rules: docs/development/validation.md
  • Release, publish, and automation flow: docs/development/release-and-automation.md
  • Agent workflow and Codex-specific guidance: this docs/codex/ section
  • Repository ownership and current gates: AGENTS.md
  • Template command and integration catalog: codex/config.toml

Status Model

  • available_now: usable in the current environment without extra setup
  • recommended_later: worth standardizing, but not assumed to be installed in-session
  • blocked_or_misconfigured: present or intended, but not currently usable from this environment

Current Environment Snapshot

Available Now

  • python is installed and is the project runtime
  • git is installed
  • the repo-local Python toolchain works through python -m ...
  • codex.exe exists through the VS Code extension install
  • openai_docs MCP is the primary recommended documentation integration for OpenAI work
  • github_mcp is worth adding later for richer GitHub context once auth and setup are stable
  • a future repo-specific skill can help with phased modernization work

Blocked Or Misconfigured

  • gh is installed, but the current GitHub auth token is invalid
  • docker is installed, but the daemon or config are not currently usable from this session
  • no MCP servers are visible in this session right now

Why This Exists

This repository uses Codex as an implementation partner during a phased modernization. These notes keep agent roles, validation commands, and optional integrations explicit so each checkpoint stays reviewable and repeatable.