Openai Codex Cli Agents.md Instructions Best Practices 2026 Prompts

<context>
You are working in an existing production project. First inspect the repository structure, package scripts, routing, data models, and existing UI patterns before editing.
</context>
<task>
Implement a focused openai codex cli agents.md instructions best practices 2026 workflow that is useful on the first screen, includes clear empty/loading/error states, and follows the project's current architecture.
</task>
<constraints>
Do not add new dependencies unless the existing stack cannot solve the task. Do not change unrelated files. Do not fake analytics, traffic, users, revenue, or performance data. Keep admin/reporting data real.
</constraints>
<acceptance_criteria>
The feature builds without type errors, has at least one meaningful test or validation path, and can be verified from a browser or CLI command. Summarize changed files and exact verification commands.
</acceptance_criteria>

Create the minimal file set for openai codex cli agents.md instructions best practices 2026. Include:
- a README-style usage note
- a config/rules file if the tool supports one
- one production prompt
- one QA prompt
- one rollback/checklist prompt
Output each file as: path, purpose, content. Keep the files copy-ready and avoid placeholders unless the value is genuinely project-specific.

Review the current implementation for openai codex cli agents.md instructions best practices 2026. Prioritize production risks: broken routing, missing env guards, weak auth boundaries, stale cache behavior, SEO regressions, accessibility issues, and missing tests. Return findings first with file paths and exact fixes. If there are no blockers, say so and list residual risks.