Your reads would be in FASTQ format, not FASTA, though you will possibly have a transcriptome/genome reference in FASTA format.

A typical workflow I would say is something like:

• Start with by getting some FASTQ files, have them in separate directories per sample (this is something that's done easily with terminal/bash commands, good opportunity to get familiar with that.)
• You may run a QC tool (like FastQC + MultiQC) on the raw data, determine if you need to apply trimming
• Apply trimming, re-run QC tools
• Align your reads to a reference (giving you a BAM file) using an aligner (some quantification tools like RSEM allow handling of this step implicitly). Kallisto (which I haven't really used) can perform quantification without alignment as described here (https://pachterlab.github.io/kallisto/starting.html)
• Quantify (for example, running Kallisto, or RSEM)

Most likely yes, that would all happen on the terminal. I would suggest looking up some tutorials or existing pipelines using Kallisto. This one looks like a good start: https://felixeyegithubio.readthedocs.io/en/latest/rnaseq/labs/kallisto/

People typically run FASTQ files to generate counts rather than FASTA files. Yes, QC and trimming would be done before hand. Use FastQC to look at all of your FASTQ files. It will tell you if you need to trim any adaptors or if there are any other QC issues. It has both a simple GUI and a command-line version if you only have access to a headless linux server.

Trimming can be done with any number of tools, but Trim Galore is pretty popular and easy to use (and made to work with FastQC). It is run from the command line.

See https://github.com/snakemake-workflows/rna-seq-kallisto-sleuth for a useful Snakemake workflow.