Skip to content

agents.md

Latest commit

 

History

History
58 lines (38 loc) · 2.5 KB

agents.md

File metadata and controls

58 lines (38 loc) · 2.5 KB
Raw

Rules for AI-assisted contributions

Scrum Helper accepts AI-assisted contributions, but all submissions must meet project standards. The contributor submitting the PR is fully responsible for correctness, scope, and validation.

For setup and contribution process details, see CONTRIBUTING.md.

Required rules

1) Understand and explain every change

  • You must be able to explain what changed, why it changed, and how it interacts with existing code paths (popup, service worker/background, content scripts, GitHub API, i18n, and browser-specific build outputs).
  • If you cannot explain a change, do not submit it.

2) Keep changes minimal and in scope

  • Keep PRs tightly scoped to the stated bugfix or improvement.
  • Do not include unrelated cleanup, drive-by refactors, formatting-only churn, or file moves unrelated to the task.
  • Do not modify generated outputs or unrelated files unless the task explicitly requires it.

3) No new feature without a linked issue

  • New features must reference a linked GitHub issue in the PR.
  • If no issue exists, open one first and align scope before implementation.
  • Bug fixes and maintenance changes should still include clear problem context in the PR description.

4) No blind AI output

  • Do not paste AI-generated code without reviewing and testing it.
  • Follow existing architecture, naming, structure, and error-handling patterns in nearby code.
  • Reuse existing helpers before adding new abstractions or duplicate implementations.

5) Mandatory validation before PR update

Run these commands locally after your changes and before opening/updating a PR:

npm run lint
npm run check
npm run build

Additional expectations:

  • If lint/check fails, fix issues before requesting review.
  • Reload and verify the built extension from dist/chrome and/or dist/firefox for the user-facing behavior you changed.
  • Validate edge cases relevant to your change (empty states, missing token/auth, API failures, rate limits, and large result sets where applicable).

Pull request requirements

  • Use a clear PR title and description with problem, approach, and validation evidence.
  • List the exact validation steps you ran and outcomes.
  • If maintainer feedback conflicts with an AI suggestion, follow maintainer direction.

Why these rules exist

  • Reduce maintainer overhead from speculative or low-signal PRs.
  • Prevent regressions from unverified AI-generated changes.
  • Keep the codebase consistent, reviewable, and aligned with project direction.