Spreadsheet. The first sheet contains, by column:

• The word itself, in citation form. If the conlang uses Latin/Cyrillic/Greek I keep it in the original, otherwise I romanise the writing system, as close to possible to the original*. No idea how I’d deal with a logographic system.
• Phonemic and phonetic transcriptions. Often more than one if I’m handling multiple con-dialects.
• Part of speech: noun, verb, etc.
• Declension/conjugation, if applicable. Heavily irregular words get marked “irregular” here.
• Sandhi classes, gender, stuff like this, when applicable. Basically, ways this word interferes on other words.
• Coarse translation, with one or two words.
• Notes for subtle irregularities (e.g. “genitive in -n”), and a finer translation than the above.

The next sheets include declension/conjugation tables, separated by part of speech. Heavily irregular words get handled here, as their own conjugations/declensions.

*even if this means doing counter-intuitive stuff, like writing ⟨b̃⟩ for /m/ because it’s how the native script handles it.

I think I’d already written a few texts in my conlang when I started compiling the dictionary, so that’s where I started, by collecting the words from all the extant texts. I also turned to older conlangs I’d made and nabbed any words from them that I thought looked good. After that, I wanted the first order of business to be translating the Swadesh 207, translating important specialized terms from my conworld, and developing numerals and kinship terms, and creating some more grammatical particles and affixes.

My plan has been to translate Wikipedia’s Basic English Combined Wordlist, but I’ve found that just going through a list of words and coining translations for them one by one is a bit tedious, so what I’ve ended up doing instead has mostly been to translate texts (song lyrics, PSAs, Wikipedia articles, titles and plot descriptions for movies and TV shows, memes/copypastas, etc) and use this as a more engaging source of inspiration for new words. I do however still plan on eventually translating the entire Basic English Combined Wordlist, and after that the nimi pu/ku of Toki Pona, and then I can say that the first draft of my dictionary is done. After that I plan on writing a second draft that cleans things up and prunes a few words.

Also worth noting that I have an appendix section in my dictionary, which I mainly use for terms that don’t make sense in the conworld but are necessary for translating “real-world” texts, e.g. country names, names of holidays, specialized terms from other fictional works, the likes. Sometimes I don’t add words to my dictionary until I’ve experimented with using them a bit and decide to keep them.

My dictionary is written in LibreOffice Writer. The dictionary’s headwords are in the conlang, written in an ASCII-friendly romanization system. Words starting with A~M are fully alphabetized, and words starting with N~Z are only alphabetized by the first two or three letters, and are otherwise in random order. Each headword is followed by a spaced em dash, and then an abbreviation of the word type in parentheses, sometimes with other information like “vulgar” or “colloquial” or “poetic” or “transitive” or “passive” or “reflexive” etc. After that is basically a list of translations into English, or a full definition in English. For instance:

kopp — (f) medium-to-dark blue; (adj) of such a shade of blue; (m) a bruise; (figurative) a romantic or platonic crush or squish.

koppatt — (v, trans. or intrans.) to bruise; (transitive) to cause [d.obj.] to develop a crush on [subj]; (intransitive) to develop a crush.

Usage notes or other notes such as irregular inflections are preceded by a ※ and placed at the end of a definition. For instance:

eyn — (b) a person, a human, an individual. ※ IRREGULAR: definite follows the long plain pattern rather than short iotating pattern.

I’m also slowly but surely working on a signed conlang, but I’m sorry to say that I haven’t really done much lexicography with that lang because it’s very tedious. My best idea thus far has been to import videos of myself signing into a program called Hydrus Client and then tagging the videos with the parameters and glosses, but this ends up being a very time consuming process.