What belongs in your AI’s instruction file, and what should stay out, so the rules that matter get followed?
- Module
- 01/05
- To read
- ~9 min
A team’s instruction file had grown to hundreds of lines, and one line said never to run migrations against the shared staging database. What do you think the agent did?
In the module you make your guess first, then read what happened and why.
Only what the agent can’t infer
Small at the root, detail pushed down
Keep it current: treat it like code
“Never” rules need enforcement
Your AI’s instruction file
A spec for your own app, as a text file you download. The module also gives you the prompt for Claude Code, Codex or Antigravity, a checklist to check what it built, and recommended skills.
- Configuration as code
- Guidance versus enforcement