Ship Better Code, Faster

Appearance

AI Usage Policy: OrbitTasks Engineering ​

This policy reflects what we learned in Workshop 5. It will evolve.

When AI tools are allowed without disclosure ​

  • Autocomplete suggestions inside the IDE (e.g., Copilot, Cursor autocomplete) that you accept inline as part of writing code.
  • AI-assisted research that informs your design but doesn't produce code that ships.
  • Refactoring suggestions you fully review before applying.

When AI tools require disclosure in the PR ​

Disclose by adding the ai-assisted label and a one-line note in the PR description.

  • AI-generated code blocks larger than ~10 lines.
  • AI-generated test files, partial or whole.
  • AI-generated documentation, READMEs, or changelogs.

When AI tools require an additional reviewer ​

Beyond the disclosure above, request review from a second teammate.

  • AI-generated code in apps/api/src/services/auth.service.ts and other security-related modules.
  • AI-generated changes to CI/CD workflows.
  • AI-generated migrations or schema changes.

When AI tools are not appropriate ​

Never AI-generate these.

  • Anything that handles user-submitted secrets or PII.
  • License headers, attribution, or legal text.
  • Commit messages or PR descriptions that misrepresent who did the work.

How we keep this policy current ​

  • Reviewed quarterly. The most recent revision date is at the bottom of this file.
  • Anyone on the team can open a PR against this policy.
  • Major changes (additions or removals from any list above) require team agreement.

Why we have this policy at all ​

We use AI tools because they make the team faster. We have a policy because unmanaged AI use creates technical debt, security risk, and accountability gaps. The policy is short on purpose. It should not feel onerous.

Last reviewed: [date]Owner: [name / team]