Documentation may or may not be forthcoming for these. Some do have a usage message though. Most are short enough that you can easily read the source code to see what they do.

Also relevant: git-util pbmutil

addup is a filter that adds up a series of numbers and prints the total. addup n splits each line into fields and totals the n'th field in each line.

also is a wrapper around guessfile and menupick, described below. It guesses which files you might have meant, presents a menu to let you select which ones you actually meant, and then opens emacs buffers for those files.

autofile is a utility for the MH mail system. It scans the current folder and applies a set of rules and actions to the mail it finds, reflining all the messages mathing certain patterns into certain target folders.

badge sets the "badge" in an iTerm2 terminal. See the iTerm2 documentation for details.

copy-if-changed doesn't work yet. Don't use it.

count generates a report of the unique lines in the input, sorted by frequency of occurrence.

date-arith takes an argument list like 2014-12-23 +12, interprets +12 as “plus 12 days”, and prints the resulting date in the same format.

disp is a triviality. My .screenlayout directory contains a number of shell scripts, saved for me by arandr, for reconfiguring the screen dimensions to match various monitor setups. I use disp 1 when I want to use the laptop screen only, work-vertical when my work monitor is in its vertical configuration, and work-vertical-stand when it is also on a stand. The disp command finds one of these and runs it; it also supports a -c flag so I can use it as the backend of a shell programmable completion function.

do-over is a triviality: it runs a certain command every time you hit enter. I don't know why I like it so much.

dt is for managing the names of KDE desktops. The System Settings widget for doing this is clumsy. Basic use is dt [ -n number ] new-name. If -n is omitted, it sets the name of the current desktop.

dus is like the standard du program, but lists the sizes of files as well as the totals for each directory. dus -n sorts the output by size.

every 15 command args... runs the command every 15 seconds. Normally each run starts 15 seconds after the last one ended; every -s starts each command 15 seconds after the last one started. -v is verbose; -x tells it to exit if any of the runs fail.

f is the most useful of the bunch. f 6 is the same as awk '{print $6}', except with 80% less typing. f -1 works.

files lists the files in the current Git repository that match a pattern. Pattern semntics are fine-tuned to my preferences. For example, files foo finds all the files whose base name contains foo, but files foo*.py finds all the files whose base name begins with foo and ends with .py. With no arguments it lists all files, and it always omits gitignored files.

getstore _url_ _file_ fetches the file at the given URL and stores it in the specified file. If you omit the filename, it will be inferred from the URL.

guessfile takes a list of patterns and prints out name of files matching one of the patterns. Patterns are shell glob style, with some modifications: if the pattern contains ::, this is translated to a slash (for searching for Perl modules) and patterns are unanchored, except that if one contains a suffix like .txt it is anchored to the end. Only the base filename is matched against the pattern, unless the pattern contains a slash, in which case the full path is matched. Environment variables control the directories to be searched (default . unless overridden by the -d flag) and to be ignored (default .git). -i makes the match case-insensitive. The withsome command, inspired by Ronald Schmidt's command of the same name, is a wrapper around guessfile and menupick: withsome command pat guesses files matching pat, filters the guesses with menupick, and then runs the command on the selected files.

ifchanged watches a file using the Linux "inotify" facility and runs a command whenever the file changes.

With arguments, localtime converts the arguments from epoch time to local time and prints the conversions one per line. Without arguments, it is a filter, replacing an epoch time at the beginning of any input line to the corresponding local time.

indir runs a command in some other directory. It saves you from having to chrid there manually and then chdir back, and may save a conceptual step.

lastdl searches the $HOME/Downloads directory for the most recent downloaded file, and simply prints its name. This is so that you can easily do mv $(lastdl) . or something of that sort. The shell mis-handles filenames that contain whitespace, so lastdl will rename the file to a whitespace-less one before printing the name.

lr performs rudimentary statistical analysis of its input, including, primarily, linear regression. The input should be two columns, of X and Y values, or one column, of Y values (and the X values are implied to be 1, 2, 3, etc.). The command then prints out the means, variances, and standard deviations of the inputs, the correlation between the two columns, and the slope and intercept of the best-fit line through the data. The -lr flag suppresses everything but the linear regression results; the -mean flag suppresses everything but the means, and so forth.

menupick is a filter reads a list of items from stdin, prints a menu of the items on the terminal, repeatedly prompts the terminal for a selection of items, and prints the selected items on stdout when the prompting is over. For example: emacs $(ls | menupick).

Responses to the prompt are a series of numbers, which identify items that are added to the current selection, and numbers with prefixed ! marks, which are removed from the current selection. Any number of such items can be entered at once, sepearated by whitespace. Ranges such as '12-23' are allowed. '*' is an abbreviation for all the items on the current page.

An empty line, or a line that ends with !, terminates the prompting. If an input line contains a single ?, the list of currently-selected items will be printed before the next prompt; if it contains ?? the menu will be redisplayed, with selected items marked.

If the number of items is large, lines containing j or n move to the next page of the menu, and k or p move to the previous page.

numfilter copies standard input to standard output, but one or more of the fields are examines for numerals, which are then reformatted to human-friendly format. Use -f to select which field. Use -F to select which formatter is used. The default converts 1234 to 1.2k and similarly. Use -b to use powers of 2 instead of powers of 10.

path prints out your $PATH variable, one entry per line, so that you can iterate over the components with for or pipe them into grep or whatever. A future version will perform the opposite transformation.

pgrep is an implementation of grep in Perl that uses Perl regular expressions, implements a couple of features I use a lot, and eliminates a vast number of features I never use.

psgrep runs ps and greps the output, but leaves intact the header line that explains what the columns mean. It also omits itself and its own ps from the output.

randword selects a random line from its input, by default a dictionary file. With -n it selects more than one line. The outputs are selected equiprobably, and the input is not stored in memory. It may soon be replaced with sample (see below).

reso prints a list of the available monitor resolutions, grouped by aspect ratio, so that you can select a resolution that has the same aspect ratio as the current one.

rmdups removes duplicate files from one or more directories. Currently, it only removes with names of the form foo (1).bar, and only if there is an identical file named foo.bar in one of the directories.

runN runs a command repeatedly with different arguments, by default on process at a time, but the -n N option tells it to run up to N jobs in parallel. This version, by Aaron Crane, has a better interface than my original implementation.

sample selects and prints a random sample of lines from the input. sample 0.23 and sample 23% select each line with probability of 23%. (They do not guarantee to emit exactly 23% of all lines.) sample 23 selects and emits exactly 23 lines, using the standard reservoir sampling algorithm.

shutup command args... runs the comand with its standard output and standard error directed to /dev/null. If the command wants to communicate with the user, it must do so by another method. shutup does not detach the terminal, so the subcommand is free to redirect its output to /dev/tty if it likes

sort-natural sorts its input lines like sort, but arranges to sort the numeric parts of the lines numerically. For example,

    a a1 a10 a100 a10x a2 a21 a3 ab b

sorts into the order

    a a1 a2 a3 a10 a10x a21 a100 ab b

with the numeric parts of the a sequence in numerical order.

sortbylength is a filter that sorts its input lines from shortest to longest.

spot reads an input and emphasizes the differences from line to line by erasing the parts of each line that match the previous line. Try ls -l | spot to see what it does.

stopafter 12 command args... runs the specified command with the specified args, but sets an alarm clock to go off after 12 seconds, which will interrupt the process (unless the command reschedules the alarm or handles or ignores the alarm signal.)

stouch adjusts the last-modified dates on a set of files so that they appear to have been written in the order given on the command line. This is useful for dealing with stupid photo gallery programs that insist on displaying files in chronological order.

suf removes the suffix of a filename and replaces it with a different suffix. For example, mv $i $(suf $i .jpg) moves img.gif to img.jpg. People usually do this with basename but suf is shorter and easier.

with-memory-limit runs a command with its memory limited by the rlimit facility.

z command args... tries to run the command on the specified arguments, but if any of those arguments appear to be compressed files, it creates a pipe, runs gzip -cd to decompress the file into the pipe, and attaches the command to the pipe instead. So for example

    perl -lne 'print if /octopus/' *

doesn't work if some of the files in the current directory are compressed, but

    z perl -lne 'print if /octopus/' *

does.