A Codex Prompt to Audit All Your Skill Files and AGENTS.md Against Six Best Practices

dr_cintas · x · 2026-09-07

The author shares a written Codex prompt that walks through every skill file and AGENTS.md in a project, checking each against six best practices derived from OpenAI's codex dx review for GPT-6 Astra—reporting only, no rewrites. Checks include: bloated or vague skill descriptions (long ones risk truncation, vague ones get misselected), missing progressive disclosure (single long files covering multiple workflows instead of a short router with on-demand docs), and rigid step-by-step recipes instead of trusting model judgment. A directly reusable audit workflow for agent engineering.

Related event: A Single Codex Prompt to Audit All Your Skills and AGENTS.md Files(2 posts)→

Original post →

More from coding & agent

coding & agent channel →