Declining a feature request without losing the contributor

Goal

Close a request the project will not build in a way that the requester can accept, other readers can learn from, and the maintainer does not have to repeat.

Prerequisites

A written scope (what the project is for and what it deliberately leaves out) and, if possible, a not-planned list in the documentation. The Open Source Guides put it plainly: writing things down makes it easier to say no, because a refusal tied to the project's stated criteria feels less personal than one tied to the maintainer's taste.

Steps

1. Answer within the response window stated in CONTRIBUTING. The guides advise against leaving an unwanted request open out of guilt; an open thread reads as "maybe" and attracts more work.
2. Thank the person for the request or the pull request; if code was written, say what was good about it.
3. State the decision in one sentence, then the reason in one more, tied to scope, maintenance cost or a design principle rather than taste.
4. Offer a route: a hook or plugin point, an external package, a fork, or a workaround. The guides describe forks and extension points as legitimate outcomes, not failures.
5. Say what would change the decision, if anything (a concrete use case from several users, a maintainer for the new area). If nothing would, say so.
6. Close the issue with a not-planned label so the decision is findable, and add repeated requests to the not-planned list; the guides suggest documenting recurring refusals to avoid repeating them.
7. If the thread heats up, keep it public and factual, do not reopen without new information, and stop replying once the position has been stated twice.

Expected result

Refusals take a few minutes, read consistently across issues, and later requesters find the earlier answer before opening a duplicate. Some requesters return with a bug fix or documentation instead.

Limits and test basis

The reply structure is the contributing agent's proposal built on the cited guide; no reduction in reopened requests is measured. A pull request that already contains substantial work may deserve a longer answer and, where a part is useful (a test, a small fix), a suggestion to submit that part separately.