The purpose of this script is to analyze plain text files (with or without the ".txt" extension) looking for broken paragraphs, i.e., paragraphs splited in more than one line, and join them in a single very long line. It's markdown friendly.

paragrapher.sh

#!/bin/zsh
################################################################################
# PARAGRAPHER revision 20
#     By Daniel "Nerun" Rodrigues
#     May 17, 2025
#     https://gist.github.com/nerun/8318924aa35f3f27231f86468804cc8c
#
# The purpose of this script is to analyze plain text files (with or without the
# ".txt" extension) looking for broken paragraphs, i.e., paragraphs splited in
# more than one line, and join them in a single very long line.
#
# This program is free software; you can redistribute it and/or modify it under
# the terms of the Creative Commons Zero 1.0 Universal (CC0 1.0) Public Domain
# Dedication (https://creativecommons.org/publicdomain/zero/1.0/).
################################################################################
if [[ -z "$*" ]] || [[ "$1" = "-h" ]] || [[ "$1" = "--help" ]]; then
    cat << EOF
Paragrapher processes plain text files to fix broken paragraphs — lines split
where they shouldn’t be — and joins them into single, long lines.

Usage: paragrapher [OPTION] <FILE>

  -c, --columns    Set minimum line length to detect paragraphs. Default is 72
                   columns. Use lower values (50 or 60) for shorter paragraphs.
                   Example: paragrapher -c 60 filename

  -h, --help       Show this help message.

  -v, --version    Show version, author, URL, and license info.

The output is always saved as filename_paragraphed.

Works well with Markdown — recognizes headings and lists. Run it directly on
".md" files: paragrapher filename.md.
EOF
    exit 0
elif [[ "$1" = "-v" || "$1" = "--version" ]]; then
    cat << EOF
PARAGRAPHER revision 20 (May 17th, 2025)
By Daniel "Nerun" Rodrigues
https://gist.github.com/nerun/8318924aa35f3f27231f86468804cc8c

This program is free software; you can redistribute it and/or modify it under
the terms of the Creative Commons Zero 1.0 Universal (CC0 1.0) Public Domain
Dedication (https://creativecommons.org/publicdomain/zero/1.0/).
EOF
    exit 0
elif [[ "$1" = "-c" || "$1" = "--columns" ]]; then
    if [[ -n "$2" && "$2" == <-> && -n "$3" ]]; then
        columns=$2
        filename="$3"
    else
        echo "Error: incorrect usage. Example: paragrapher -c 72 file.txt"
        exit 1
    fi
else
    columns=72
    filename="$1"
fi

if [[ -z "$filename" ]]; then
    echo "Error: no input file specified."
    exit 1
fi

# Check if file exists
if [[ ! -f "$filename" ]]; then
    echo "Error: there's no file \"$filename\"."
    exit 1
else
    # Check if file is a plain text file
    file_test=$(file -b "$filename")
    echo "$file_test"

    if [[ "$file_test" != *"text"* ]]; then
        echo "Error: \"$filename\" is not a plain text file."
        echo "(detected: \"$file_test\")."
        exit 1
    else
        # Check if it is UTF-8
        if [[ "$file_test" != *"UTF-8"* || "$file_test" == *"with CRLF line terminators"* ]]; then
            if command -v dos2unix >/dev/null 2>&1; then
                dos2unix "$filename"
            else
                echo "Consider installing \"dos2unix\"."
            fi
        fi
    fi
fi

# Verify filename extension, if any
if [[ "$filename" == *"."* ]]; then
    ext="${filename##*.}"
    ext="${ext:l}"
    output="${filename%.*}_paragraphed.$ext"
else # filename has no extension
    ext="txt"
    output="${filename}_paragraphed.$ext"
fi

# Add an empty line to the end of the file to avoid paragrapher not reading the
# last line
last_line=$(tail -n 1 "$filename")

if [[ -n "$last_line" ]]; then
    echo "" >> "$filename"
fi

# Create or clean the output file
:> $output

paragraph=''

_dump_paragraph() {
    if [[ -n "$paragraph" ]]; then
        echo "$paragraph" >> "$output"
        echo "" >> "$output"
        paragraph=''
    fi
}

# https://mywiki.wooledge.org/BashFAQ/001#Trimming
# while IFS= read -r line; do
while read -r line; do # reads file filename, then, for each line it does:
    size=${#line} # get the length of the line

    # if the paragraph is empty, and in the first iteration it will be, then do
    if [[ -z "$paragraph" ]]; then
        # if the line length is greater than the columns variable
        if [[ $size -ge $columns ]]; then
            paragraph="$line"
        else # 0 <= size <= 71
           echo "$line" >> $output
        fi
    # is not the 1st iteration, paragraph already has content
    else
        # If the last character of paragraph does not match a period (.) or a
        # colon (:), execute the block.
        last_char="${paragraph: -1}"

        if [[ "$last_char" != "." && "$last_char" != ":" ]]; then
            if [[ "$ext" = "md" ]]; then
                case "${line:0:7}" in
                    \#\ * | \
                    \#\#\ * | \
                    \#\#\#\ * | \
                    \#\#\#\#\ * | \
                    \#\#\#\#\#\ * | \
                    \#\#\#\#\#\#\ * | \
                    "•"[$'\t ']* | \
                    "-"[$'\t ']* )
                        _dump_paragraph

                        if [[ "${line:0:1}" = "#" ]]; then
                            echo "$line" >> $output
                            echo "" >> $output
                        else
                            paragraph=''
                            paragraph+="$line"
                        fi
                        ;;
                    * )
                        paragraph+=" $line"
                        ;;
                esac
            else
                paragraph+=" $line"
            fi
        else
            _dump_paragraph

            # if the line length is greater than the columns variable
            if [[ $size -ge $columns ]]; then
                # concatenate the new line with the previous ones, creating a
                # paragraph
                paragraph+="$line"
            else # 0 <= size <= 71
                # if not greater, just write it to file
                if [[ $size -ne 0 ]]; then
                   echo "$line" >> $output
                fi
            fi
        fi
    fi
done < "$filename"

# Clean output file
sed -ri "s/’|‘/'/g" $output
sed -ri "s/“|”/\"/g" $output
sed -i s/$'\u000c'//g $output # Form feed (U+000C)
sed -i "s/ -- / – /g" $output # En Dash (U+2013)
sed -i '/ --[[:digit:]]/ s/--/–/g' $output # En Dash (U+2013) representing a minus sign
if [[ "$ext" = "md" ]]; then
    sed -i 's/^[•-] [ \t]*/  - /g' $output # Bullet (U+2022) or list
fi
echo "" >> $output # i like to end files with empty line

It is fully compatible with BASH 5+. Just change the shebang to #!/usr/bin/env bash