The temptation to stretch is strongest exactly when it should be resisted, which is during a quiet month. Work arriving when the calendar is empty looks considerably more appealing than the same work would in a busy period, and the judgement about whether you can actually do it is worst at that moment.
Be direct rather than vague. Saying that something is outside what you do is a complete answer, and it is better received than a hedged response that leaves somebody uncertain whether to pursue it. Vagueness produces a follow up, which means having the conversation twice.
Refer where you genuinely can, since that converts a disappointing interaction into a useful one. Knowing who does good work at the edges of what you do is part of being useful, and an introduction costs nothing and is remembered longer than the refusal. Where you have no suggestion, saying so plainly beats inventing a direction.
Consider whether it is adjacent rather than outside, which is the genuinely difficult category. Work you could do competently but have not done before is a different situation from work you have no business attempting, and the honest response is to say that, explain what it would involve, and let them decide with the information.
Never learn at a client's expense without telling them, which is the line that matters. Taking on unfamiliar work while implying it is routine is how businesses produce bad outcomes and lose their reputation for the categories they are actually good at. Disclosed, the same arrangement is sometimes perfectly acceptable to a client who values the relationship.
Watch the pattern in what people ask for, because repeated requests for something you do not offer is market information. It may indicate a genuine adjacent service worth developing, or it may indicate that your description is attracting the wrong expectations. Both are worth knowing and neither is visible from a single conversation.
Keep the door open where the fit might change. Somebody who needed something you do not offer may need something you do in six months, and ending the conversation warmly rather than transactionally is what makes that possible. A short note saying what you do handle, in case it comes up, is enough.
Then resist expanding your services because one person asked. A first year business accumulates offerings this way, each justified by a single request, and ends the year doing a collection of things nobody chose. Adding a service should follow a pattern of demand rather than an individual conversation.
Answer promptly rather than deliberating, since a same day no is a courtesy and a delayed one is a cost you imposed. The person can approach somebody else while the need is current.
Build a short list of people you would refer to, before you need it, since the useful referral requires knowing somebody rather than searching at the moment somebody asks. Three or four names across the adjacent categories covers most of what arrives.
Record what you declined and why, briefly, since across a year that list is the clearest description of your actual positioning and it frequently reveals a category arriving often enough to be worth serving deliberately.
Consider a partnership rather than a referral where the request recurs, since an arrangement with somebody who handles the adjacent work lets you keep the relationship while they do the part you do not.
Be careful about recommending anybody you have not seen work, since a referral carries your judgement and a poor outcome reflects on you rather than on them.