Stop fixing your AI agents in chat

Description and use

Moving from chat corrections to shared Markdown rule files provides a single source of truth, version history, and reviewability, eliminating the need for chat‑based fixes.

Problem

Chat corrections to AI agents are transient and unreviewable, causing repeated mistakes and lack of traceability.

Outcome

Persisting rules in shared Markdown files provides a single source of truth, version history, and reviewability, eliminating the need for chat‑based corrections.

Workflow

• 1. Identify the incorrect behavior in chat.
• 2. Ask the agent which file should contain the correct rule.
• 3. Edit the identified Markdown file to add or modify the rule.
• 4. Commit the change to the repository.
• 5. Run the agent again; it reads the updated file and applies the new rule.

Evidence

A correction in chat lives in one conversation. When that session ends, so does the rule.

Source

Commit it like any other change. The next run, in your session or a teammate's, uses the new rule.

Source

You get three things chat never gave you: One source. History. Review.

Source

We keep the agents' knowledge as plain Markdown files and let the agent search them, instead of running them through an embedding index first.

Source