Track intent and vet plans before running code

A developer records user messages for later review and has agents inspect written plans against project rules before work starts.

Done with Custom agent· Intermediate· Oct 2026

What they did
They keep a running log of each user message so earlier direction is not lost if a session is compacted or restarted, then revisit that log to pull out the latest version of the intent. Before they begin a substantial task, they draft a plan, save it, and ask an agent to compare it with the project’s limits.
How it went
No outcome or measurements are given. The commenter treats both habits as safeguards for staying aligned with user intent, and says trust in agent-generated code depends on the project itself.
Tips
This works best when your project has clear constraints and you want a record of why decisions were made. Make sure the agent pauses for your approval before any action that could have lasting effects.

Try it yourself with Claude Code

Act as my planning assistant for [project]. Each time I send a message, add it to [log file path] so my intent is preserved. Before you do any work, write your proposed approach in [plans folder]/[plan file], then check it against these rules: [constraint 1], [constraint 2], [constraint 3]. Present the plan and your review, then stop and ask for my approval before changing files, sending anything, deleting anything, or making any payment. After I approve, complete the task. When you finish, summarize the intent you inferred from the log and mention any changes made to the original plan.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on news.ycombinator.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.