Codex Ecosystem Setup

Codex Ecosystem Setup Skill

This skill handles bootstrap and MCP management for Codex projects.

Bootstrap policy

Do not create or modify .codex/ automatically on plugin activation.

Use this policy:

1. Check whether .codex/config.toml exists.
2. If missing, explain what will be created.
3. Ask for confirmation.
4. Create files only after confirmation.

Bootstrap actions

When approved by the user:

mkdir -p .codex
cp "${CLAUDE_PLUGIN_ROOT}/.codex/config.template.toml" .codex/config.toml
cp "${CLAUDE_PLUGIN_ROOT}/.codex/recommended-mcps.toml" .codex/recommended-mcps.toml

If plugin-root files are unavailable, create .codex/config.toml with safe defaults:

model = "gpt-5.4"
model_reasoning_effort = "xhigh"
sandbox_mode = "workspace-write"
approval_policy = "on-request"

Curated MCP set

Only the verified set is considered curated here:

• context7
• playwright
• postgres
• filesystem
• github

All are disabled by default in template config.

MCP enablement workflow

1. Analyze task requirements.
2. Suggest relevant MCP(s).
3. Ask before changing .codex/config.toml.
4. Enable only requested MCP(s).
5. Remind about required environment variables.

Notes

• Prefer minimal enablement.
• Use recommended-mcps.toml for server reference blocks.
• For non-curated MCPs, use /codex-mcp-search and manual user confirmation.

References

• skills/codex-integration/references/codex-prompt-guidance.md — GPT-5.4 prompt patterns
• skills/codex-integration/references/codex-cli-reference.md — CLI commands and config

Converted and distributed by TomeVault — claim your Tome and manage your conversions.