Best practices for Discord bots

NB: These guidelines are intended for bots running on public servers. If your bot is restricted to private ones, this document likely doesn't apply to you.

1. Commands should be explicitly invoked. Bots should not activate on normal chat. Instead, use a command prefix or only respond when your bot is directly @mentioned.
2. Use unique prefixes. Single-character prefixes such as !, $ and . are commonplace for activating commands and lead to overlaps with other bots. Should you opt to use a prefix for your bot, consider using words (owl) or unique Unicode characters (¨). Also, you should avoid using # or @ as prefixes since they can be used to mention a channel or a member. Ideally, your bot's prefix should be configurable on a server-by-server basis, so that the server owners can ensure every bot has its own unique prefix of their choice.
3. Don't be greedy. Restrict yourself to a small number of prefixes to reduce the risk of collision with others.
4. Don't overuse mentions. If you reply directly to a command, don't use a mention, they can lead to bot reply loops. Mentions are fine if a long-running command is executed, but private messages are a good alternative.
5. Have an info command. It should provide information about the bot such as what framework it is using and the used version, help commands and, most importantly, who made it.
6. Don't reply with "invalid command". If a user uses a command that does not exist, then let it fail silently. Do not have it reply with something like "invalid command". Though if the command is correct, but arguments are wrong then it's okay to reply with "invalid args". If there is more than one bot in a server that shares a prefix, this can result in very obnoxious usage. If your bot's prefix is configurable, this rule can probably be safely disregarded.
7. Be respectful of Discord's API. Bots that abuse and misuse the Discord API ruin things for everyone. Make sure to factor in rate-limiting and backoff in your bot code, and be intelligent about using the API. Make sure to ask for help if you're unsure about the right way to implement things.
8. Ignore both your own and other bots' messages. This helps prevent infinite self-loops and potential security exploits. Using a zero width space such as \u200B and \u180E in the beginning of each message also prevents your bot from triggering other bots' commands. The Discord API also tells you if a user is a bot (bot property on User objects - see the reference).
9. Keep NSFW features locked to NSFW channels All NSFW commands/features should only work in (Discord) NSFW-marked channels.
10. Use mentioning the bot to help users.. Allowing a mention as the prefix ("@MyBot help") or adding a way to find the bot's prefix with only a mention ("@MyBot" or "@MyBot, what's your prefix?") will help users who are new to your bot in getting started. (Make sure that whatever the message is, it's easily found. A great way to do this is by including it in your bot's presence.) The alternative is brute-forcing punctuation characters to find it, which will be difficult for bots following 2 and 3. Plus, a mention is the most unique prefix of all.

If you have an idea for an addition or change to this document, please make a pull request and we can discuss it.