Project planning

Bot Project Checklist: From Idea to Reliable Deployment

Use this before asking an AI model to generate code or before connecting a prototype to a real external service.

Friendly bot-building team working through a checklist
Illustration: clear requirements and a checklist reduce expensive debugging later.

The purpose of this checklist

Most bot problems are not caused by one missing line of code. They come from unclear inputs, hidden assumptions, missing state, unsafe retries, leaked secrets or no plan for restarts. Write these decisions down first.

1. Define the job

2. Define the data contract

3. Choose the runtime

4. Handle secrets correctly

5. Design failure behavior

6. Test in layers

  1. Test parsing and validation with local examples.
  2. Test decision logic without an external API.
  3. Mock broker/API responses.
  4. Use demo/paper environment.
  5. Test invalid input deliberately.
  6. Test network failure and timeout behavior.
  7. Restart the service and verify recovery.

7. Make the system observable

Logs should answer: what event arrived, what decision was made, why it was accepted or rejected, what external request was attempted, what response came back and whether state was updated. A heartbeat or health endpoint should tell you that the service is alive; a useful monitor should also tell you whether it is still receiving fresh data and completing its work.

8. Write an AI prompt from the checklist

Build the smallest working version of this bot.
Runtime: [exact runtime]
Language/version: [exact version]
Input schema: [exact fields and types]
Output/action: [exact behavior]
State: [database/KV/file/none]
Authentication: [mechanism, no hard-coded secrets]
Duplicate rule: [event_id behavior]
Timeout/retry rule: [exact behavior]
Logging: [fields to log]
Tests: [required tests]

Before coding, list assumptions and missing requirements.
Do not invent platform APIs or undocumented functions.