OpenSSL コマンドビルダー

How it works

openssl is not one program but roughly thirty, each with its own option set, and the manual page is organised around those programs rather than around what anyone is trying to do. The result is that “make a self-signed certificate with a SAN” lives under req, which is documented as the certificate request command, and that the flag which actually decides whether browsers accept your certificate — -addext — appears nowhere near the top. This page inverts that: 36 tasks phrased as goals, each producing a command whose every flag is spelled out underneath.

The explanations are the point. A command you copied and cannot read is a command you cannot adapt, and openssl’s flags are unusually easy to misread — -noout suppresses the object rather than all output, -nocrypt is the difference between a script that runs and one that hangs at a prompt, and -copy_extensions is why a certificate you issued yourself has no hostnames in it. Each task also lists the specific ways it goes wrong, because those are what actually costs the afternoon.

Values you type are quoted for a POSIX shell before they go into the command, so a path with spaces or a subject with a comma survives the copy-paste. If a value contains something a shell would normally act on — backticks, $( ), a semicolon — the builder says so, because the quoting means it will be passed through literally rather than executed, and that is occasionally the surprise rather than the reassurance. Windows users should note that cmd.exe and PowerShell quote differently; the commands here assume sh, bash or zsh.

The reverse mode splits a pasted command the way a shell would — respecting quotes, escapes, pipes and redirections — and looks each flag up in a dictionary covering 180 options across 24 subcommands. Nothing is executed, and a flag that is not in the dictionary is labelled unknown rather than given an invented meaning. For the authoritative list your build accepts, openssl <subcommand> -help is always right and always local.